Home

THE T@C USERS GROUP NEWSLETTER EDITOR ROBERT

1. 43 Macros on Microfiche D PEE 43 Patrick Milligan Erratum NOFILL Program 43 Patrick Milligan eter m nee 44 Brendan McKay A Macro 46 Michael Spivak Macro 4 TIE 50 Max Diaz F cil TEX omitted obsolete Appendix A Problems Lynne Price Greetings from the 54 Problems from the TEXarcana Course 54 Johnny Stovall Balancing columns of text and translation stovall Michael Spivak Input Dependent Macro Redefinition spivakprob Letters Lynne Price we a Eat ww sg ae 57 Dreamboat
2. AT THE 1981 SPRING DECUS U S SYMPOSIUM Patrick Milligan BNR Ine In keeping with a long standing tradition there was a TEX Birds of a Feather session at the recent DECUS symposium held May 18 21 in Miami In case you didn t know DECUS stands for Digital Equipment Computer Users Society The panel members were Patrick Milligan Coordinator for DEC 20 Phil Sherrod Coordinator for DEC 10 Barry Doherty American Mathematical Society Bob Phillips Oregon Software It comes as no great surprise that DEC users are interested in TEX Don Knuth s book TEX and METAFONT New Directions in Typesetting is jointly published by the American Mathematical Society and Digital Press In addition versions of TEX exist for the DEC 10 DEC 20 and VAX VMS computers Since the DECUS Symposium was held im mediately after the TEX Implementors Workshop at Stanford emphasis was placed on the installa tion of TEX Output samples from various devices including Versatec Canon LBP 10 and Alphatype were passed around A free TEX sample the TUG membership forin was distributed The content and utility of recent TUGboat issues was described In short a sales pitch was made for TEX TUG and TUGboat Another session on is planned for the next DECUS Symposium to be held December 7 11 in Los ngeles k Rachel Schwab TEX exists on the DECsystem 10 at t
3. enon This is an integer function that has the following parameters Var Var Var Var Var Var Var Var Name FYL FONTINFO FMEM WDBASE HTBASE DPBASE ICBASE LGBASE KRBASE EXTBASE PARBASE FCKSUM FPFB FSIZE FPFI FMEMPTR PSIZE ATCLAUSE Type INTEGER FNTiNFOARRAY FMEMARRAY FBASEARRAY FBASEARRAY FBASEARRAY FBASEARRAY FBASEARRAY FBASEARRAY FBASEARRAY FBASEARRAY FDASEARRAY FBASEARRAY FSIZEARRAY FPIARRAY INTEGER REAL BOOLEAN Reads font information from file FONTFIL The integr FYL is used as index the various array parameters to establish the destination of this information RELEASE Procedure with one parameter Name INTECER The integer FYL must be in the range 1 6 It selects one of ICHAN1 ree ICHANG and executes RESET ICHANXx followed by MILPTR FILPTR 1 This closes and releases the indicated file and frees the entry in FILENAME RSETFILE Procedure with four parameters Name ID FNAME FDIRECTORY FDEVICE Type INTEGER CHAR9 INTEGER CHARS intoger must be in the range 2 6 It selects one of ICHANI through 6 to be opened for input and associates it with FNAME FDIRECTORY and FDEVICE TUGboat Volume 2 No 2 45 Examples DefineFont emtt tt typewriter font DefineFont mflogo mf logo METAFONT logo font def Def ineFont 1 2 Mf UserFonts xdef 2 send9 Error No font
4. 25 Summary of Computing Equipment and Output Devices 25 Luis Trabb Pardo Imagen Canon LBP 10 26 Site Reports CDC Cyber Thea Hodge and Michael Frisch TEX Under the North Star 28 DEC 10 and 20 Phil Sherrod DECSystem 10 20 Implementation Workshop Announcement 28 Barbara Beeton AMS Site Report 29 Patrick Milligan TEX at the 1981 Spring DECUS U S Symposium 29 Rachel Schwab at 29 Univac 1100 Bill Kelly An Implementation Report for the Univac 1100 30 VAX Monte Nichols Barry Smith Availability of Oregon Software Implementation of TEX for the VAX VMS 33 FAKES Continued 4444 TUGboat Volume 2 2 Contents Continued Fonts Barry W Doherty TUG Font 34 Thomas Hickey The Status of METAFONT at 35 William J LeVeque Font Development at the American Mathematical Society 39 Ronald Whitney 40 Barry Doherty and Ronald Whitney Proofmode and Magnificstion 40 Warnings amp Limitations Barbara Beeton Uppercase Update Fickle 42 Macros Lynne Price Greetings from the
5. or B Since font definitions are revoked at the end of groups the total effect is nothing useful It is sometimes handy to have another version of Mf which avoids this rather unsatisfactory state of affairs While we re at it we ll change the format to If char char2 true text else false text endif a few fewer braces never hurt anybody Three possible definitions for If are as follows 1 der 12 1 2 3 else 4 endif ir 142 gdet Iftemp 3 else gdet Iftemp 4 Iftemp 2 def else i endit def endit det Ir 1 2 it 1 2 gdef Iftemp else gdef Iftemps i eise Iftemp 3 der Ir 1 2 it 1 2 gdef Iftemp t else 2 endif 1 else gdef Iftemp at else 2 endif 2 Iftemp All three definitions work in most ordinary cir cumstances The first definition has the unpleasant peculiarity that any s which occur in true text or false text must be typed as a prob lem which grows exponentially if Ifs are nested The second definition avoids this problem but has another deficiency it won t nest properly why The third definition avoids both problems If true text or false text contains another If simply enclose it in s This doesn t cause grouping of course but it will ensure that each else or endif gets paired with the right I 3 Recursion Although the manual apparently never says so the macro facility in TEX is completely recursive In
6. gives a different error message Input page ended on nesting level 1 but the basic problem and the solution is exactly the same If we make the definition def If 1 then 2 else 3 if 1 gdef result 2 else gdef result 3 result then we can safely use constructions like Mf T flag then else The token then is made part of the syntax of If so that we can have constructions like If cs a cs b then where cs 1 is a control sequence with one argument b Although char and char2 may be specified by control sequences like flag which TEX ex pands out they cannot involve 1 1 again Suppose for example that we have already defined def ab 1 if 1a T else if 1b T else F so that ab 1 is T if 1 is a or b and F otherwise We would now like to define cs 1 to be true text if 1 is a or b and false text otherwise We cannot conveniently define def cs 1 if T ab i true text Velse false text If we do this then the input cs x will become Mf T if xa T else if xb T else F true text Velse false text which causes TEX to try to compare T with if giving an error message Of course the test for 1 being a or b could be made part of the definition of cs but the following scheme is far more advantageous def ab 1 if 1a gdef Ab T else it ib gdef Ab T else gdef Ab F def cs i ab i Mf TMb true text Velse false text In i
7. Son of TEX Even before TEX s final release there have been numerous suggestions for what might or ought to do The spirit of these modifications is to allow more specialized typeset ting to be done without damaging the compatibility with standard TEX Some desired features include a more suitable input language more tractable error messages incorporation of graphics output non English hyphenation capabilities batch mode rather than interactive processing and real time interactive TEX See the article by Price p 58 1 00 2 00 Macro packages Already several major macro packages have been developed see the documentation on the macro packages by Keller and D az for instance as well as Spivak s AMS TEX in various issues of TUGboat Here there was an at tempt to focus on standards and conventions of sible interest to macro writers questions of com patibility consistency in font naming conventions for replacing characters found on the Stanford non standard terminal keyboards See the articles by Milligan 44 and Price 43 this issue 2 00 3 00 Output devices and thetr interfaces A somewhat more specific examination of some of the more common output devices their characteristics and what is required of their interfaces 3 00 4 00 Architecture sessions About half dozen groups formed to discuss their particular problems Major sessions included IBM VAX DEC
8. def cm 1 2 endd ismarker 2 Mf T ans gdef nextcm else advcounti gdef nextcm cm 2 endd nextcm def countcommas i setcounti 0 cu 1 marker endd 54 Then countcomnas 1 makes the value of counti be the number of commas in 1 The endd trick is used to handle hidden commas but the trick isn t needed since we don t care what Xem does to 1 d If we do countcommas 1 then ifpos1 will tell us whether amp 1 contains at least one comma But it is preferable to use the following scheme which doesn t involve any counters and which stops as Soon as the first comma is found def cm 1 2 ismarker 2 Mf T ans gdef neztcm else gdef Hascomma T gdef nextcm 1 marker nextcm def hascomma 1 gdef Hascomma F cn 1 marker e Suppose we want to perform the operation in part b on some input of the form string string string where the separator is the control sequence which is never used in isolation and is initially defined by def We could use exactly the same scheme replacing def op 1 2 endd by def op i 2 endd But we can also take ad vantage of the fact that the separator is a control sequence to obtain a definition that is both more elegant and more efficient def op 1 A kil1 1Z def operate 1 def op 6p 1 def op kill The def op needs to be replaced by gdef op if op puts things inside braces in thi
9. 10s and 20s CDCs 4 00 5 30 Output device demonstrations This was devoted to Trabb Pardo s presentation of the Canon Laser Printer see his article in this issue p 26 and a tour of BNR given by Milligan equipment including a Versatec PERQ and Alphatype k koX ks ks kk k k Attendees TEX Implementors Workshop Stanford May 14 15 1981 Adamo Vincent Texas A amp M University Amabile Carolyn National Information Systems Ash William Stanford Linear Accelerator Center Ball George Washington State University Beebe Nelson University of Utah Beeton Barbara American Mathematical Society Bennison John Brown University Berns Eagle Stanford University Blair John CALMA Broadwell Peter Univ of California Santa Cruz Brown Malcolm Stanford University Buckle Normand University of Montreal Bupara Sarge Exxon Office Systems Carnes Lance Gentry Incorporated Chaffee Roger Stanford Linear Accelerator Center Conley Marsha University of Illinois Copeland John Cralle Robert Lawrence Livermore Lab Dailey William H Letterman Army Institute Day Christopher Lawrence Berkeley Lab Diaz Max Stanford University Doherty Barry American Mathematical Society Doob Michael University of Manitoba Durling Bob University of California Santa Crus Faul Don Lawrence Livermore Lab Faulkner Thomas Washington State University Forster Doug Stanford Un
10. 140 Singer Wang 1 8400 Facom 230 75 M 180 990 Foonly TRS 80 Fujitsu M190 Tandem HP 1000 Telefunken TR440 HP 3000 seUnivac 1300 Honeywell Univac 90 IBM Univac System 80 303X VAX IBM 3081 UNIX BM 360 VMS 370 Wang 2200 IBM 43XX Wang OIS IBM Series 1 Xerox Alto ICL 1904 Xerox Sigma ICL 2960 2980 Z80 Intel 8080 28080 Interdata OUTPUT DEVICES AM Comp Set 4510 AM CompEdit Alphakey Multisetter seAlphatype CRS Anadex Autologic APS 5 Bobst Eurocat CalComp Canon LBP Compugraphic Compugraphic 7500 eCompugraphic 8600 Compugraphic Editwriter Compugraphic Unisetter Compugraphic Videosetter Diablo Dicomed D47 F lorida Data GSI GSI phototypesetter Graphiset 8 Harris 7400 Harris phototypesetter Hell Digiset 3800 IBM 6670 IBM laser printer COMp86 HI VideoComp laser Mergenthaler Mergenthaler CRTronic Mergenthaler Linotron Mergenthaler Omni Mergenthaler VIP NEC Spinwriter Nortext typesetting system Olivetti Photon Pacesetter Printronix QUME Sanders Tektronix Trilog C 100 eVarian es Versatec Wang phototypesetter Xerox Xerox 1700 Xerox 5700 Xerox 9700 9Xerox Dover vsXerox 26 TUGboat Volume 2 No 2 IMAGEN CORPORATION page 1 of 2 Description of IMAGEN products INTELLIGENT PRINTER SYSTEM IMPRINT 10 A xerographic intelligent printer system developed and introduced by the IMAGEN Corporation with initial deli
11. Now if a job specifies font A CMR10 at 15pt say that font s magnification is recorded as 1500 in its font definition When a spooler reads this DVI file it will try to use the file CMR10 150VNT or CMR10 150ANT depending on the device which is just like CMR10 100VNT but the dimensions of all its characters were multiplied by 1 5 before they were digitized An uppercase W 10 is 10pt wide but CMR10 at 15pt has a 15pt wide W so after 87 is seen horizontal coordinate is in creased by 15pt X 254000rsu 72 27pt Overall magnification is taken into account after all other calculations for example at magnification 1200 the font CMR10 120VNT would be used Note that if the user had asked for 10 at 15truept the factors would cancel out so that CMR10 150VNT would be the font chosen regardless of magnification The magnification factor is given times 100 in the font file name so that roundoff error due to several mul tiplications will not affect the search for a font with characters of the right size This convention about font file names is merely a suggestion of course it is Dot part of the DVI format se TUGboat Volume 2 No 2 c5 6 Appendix Comparison between version 0 and version 1 Note that DVI files have an ID byte at the end of the postamble which tells what version they are The changes since version 0 are DVI files now use the upper bit
12. TUGboat Volume 2 No 2 are going to be quite different One of the prob lems with our last meeting was its undifferentiated nature Although almost everyone got quite a bit out of some particular talk or meeting it wasn t easy to know beforehand which one it would be This is probably only to be expected at the initial stages especially since so many different levels of are being addressed but with Tom s help the Cincinnati meeting ought to be better struc tured so that people can know what will be useful to them and what can be skipped Perhaps we ll even be so organized that we can propose the organization for the next meeting Let s hope so REPORT ON THE TUG STEERING COMMITTEE MEETING Robert Morris The TUG Steering Committee meeting took place in two sessions The first on May 13 simply set the loose agenda for the second which was a public meeting on the evening of May 14 The following actions were taken a few of these may have been taken at tbe loosely organized general membership meeting on May 15 By acclamation Mike Spivak was declared Chair of the Steering Committee Richard Palais will be out of the country for year b Personal dues will be raised to 15 for 1982 but no institutional dues are contemplated pending TUG offering something to its members beyond the newsletter c The Treasurer s report was approved a version updated through June 30 appears on page 5 In summary
13. available in Pascal insofar as we know The macro code serves well as structured documentation but it is difficult to refer back and forth between the macro code and the Pascal code This is because the macro code is divided into sub sections in a struc tured which expand into linear Pascal code Also the use of macros to provide alternate names for variables in the macro language means that single Pascal statement may look entirely different in the two listings Another drawback is that expressions involving constants are folded into single constant This makes compilation efficient but changing array sizes difficult We ran into this 31 problem in trying to reduce the size of the array to meet a memory restriction in our compiler The ultimate solution is to distribute UNDOC with TEX But admittedly this work should take second priority to work on itself A difficulty which may be impossible to surmount with this approach is that the output from UNDOC may still be un suitable for compilation depending on the features of the compiler For us this came about because of the environment feature of UW Pascal other sites have had to hand code the default case of the Pascal CASE statement If this situation occurs it would be impractical to correct errors in the macro source and all fixes will end up being done in the Pascal code anyway It is unclear whether it is realistic to expect UNDOC t
14. comma or Ncomma This means that the 1 appearing in check 1 endd is empty or and thus the 1 in empty 1 tricka tricka trickb is empty ii If 1 is a group that happens to begin with a comma then comaa 1 will define TUGboat Volume 2 No 2 ans to be T whereas we want it to be F this admittedly is a matter of taste So we will use emptygp and single to check on these possibilities def comma i emptygp 1 endd Mf T ans gdef ans F else single i endd if F ans else check 1 endd Then comma 1 defines ans to be T if 1 is or and F otherwise The inability to distinguish between and isa minor problem that seems insurmountable AMS TEX needs many such checks so they all made in terms of one generalized check For example comma is actually defined by def comma i compare 1 where compare is defined as def compare 1 2 def check 1 1 2 empty 1 tricka tricka trickb emptygp 2 endd Mf T ans gdef ans F else single 2 endd if F ans else check 2 1 endd The was made part of the syntax for compare to allow def space i compare j 1 Saving Braces We have just seen that there can sometimes be problems when braces are removed from the argu ment of a control sequence Actually the prob lem can be much more critical For example the AMS TEX control sequence dots 1 first examines 1 to determine what sort of do
15. s all about 3 MULTIF PROCESS APPROACH It was clear from the size of the SAIL version that could not possibly exist as a single process within a 16 bit address space Because UNIX does not support a runtime overlay system we decided to split into two separate but concurrent processes We decided to use UNIX s simple but elegant system of pipes The UNIX pipe is mechanism by which the output of one process pass1 is routed to the input of another process pass2 One early question concerned whether there existed a division point TEX where this simple tandem scheme could be implemented or were feedback paths from pass2 to pass1 always necessary Figure 1 is a simplified block diagram of data flow The vertical dashed line shows the most obvious place of division The balancing of instruction and data space requirements between the two passes along with numerous other details also affected this division point After examination of the boxes labeled Main Control and State Stacks was determined feedback paths would unfortunately be required Passi and pass had to exist as processes coupled by two pipes one for communication in the forward direction and one for the reverse direction few words about virtual memory techniques are in order much data both predefined and generated as possible was to be kept memory resident for obvious performance reasons Since a 16 bit data space was simply n
16. the individual membership fees and excess workshop revenue will cover the publi cation of TUGboat and minor administrative expenses for this year d The idea of having architecture specific im plementors workshops preferably at suc cessful site was endorsed These would be highly technical and financially self supporting Vanderbilt may organize one for TENEX sites see page 28 in this issue for an announcement e tape standards committee was established to propose formats for the exchange of files first proposal is put forth by Patrick Milligan on page 10 f Lynne Price agreed to edit the and Problems columns in TUGboat and to serve TUGboat Volume 2 2 focal point for discussion on the next generation of with emphasis on user friendliness g It was agreed to call a general membership meeting to coincide approximately with the winter meeting of the American Mathematical Society next January in Cincinnati see the preliminary announcement by Tom Pierce on page 8 h Don Knuth announced his desire intention to have fully frozen by the end of the year and to publish the theory and workings of TEX early in 1982 It was decided that the architecture coordina tors should not in general be those actually im plementing in order to shield the implementors from repetitive questions This has worked well for the VAX VMS implementation and will be gradually accomplishe
17. 5 1026 West Maude Sute 309 Suenyvale 94068 408 733 6617 HICKEY Thomes OCLC TUGboat Volume 2 No 2 ADDRESSES OF OFFICERS AUTHORS AND OTHERS HOWERTON Charles P National Bureau of Standards Boulder Labs room 3562 7 325 South Brosdway Boulder CO 80303 303 499 1000 ext 4433 KELLER Arthur Department of Computer Science Stantord University Stanford CA 94305 415 497 3227 KELLY Academic Computing Center Unaversity of Wisconsin Madison 1210 W Dayton Street Madison Wi 53706 608 262 8821 KNUTH Donald Department of Computer Science Stanford Unversiy Stanford CA 94305 LeVEQUE Witham 3 Armencan Mathematical Socety 6248 Providence Ri 02940 401 272 9500 LIANG Frank Department of Computer Science Stanford Uraversity Stanford CA 94305 MeCLURE Robert 5 1026 West Maude Suite 309 Sunnyvale CA 94068 406 733 6617 McCOURT Scott Burroughs Corporation ECD Project Corporate Drive Commerce Park Danbury CT 06810 203 194 619 McKAY Brenden D Vanderbit University Computer Science Department Bor 70 Station Nashvdle TN 37235 615 322 6517 MILLIGAN Patrick 415 969 9170 ext 2897 MORRIS Robert Mathematics Departement UMASS at Boston Boston MA 02125 617 287 1900 ext 2545 WICHOLS Monte Exploratory Chemistry Divison Sandia National Laboratory 8313 Lwermore 94550 415 422 2905 PALAIS Richard 5 Depa
18. 500 Note that I have included all commercial or ganizations and all users of TEX who use it to produce publications for sale e g the AMS and university presses as one class of users I hope the precise figures and the ratios will be the subject of much discussion in this forum because I will ask for formal ratification of some such struc ture at the annual meeting in January I will collate any replies this note brings Please mail them to me at before Sept 1 IBM Cambridge Scientific Center 545 Technology Square Cambridge MA 02139 after Sept 1 Dept of Mathematical Sciences UMASS Boston Boston MA 02125 Arpanet address ramOmit me TUGbost Volume 2 No 2 you are especially anxious that the full text of your reply be published in TUGboat please so indicate Editor s note X3J6 meeting described above has been vescheduled and Lynne Price wil probably attend rather than Mike Spivak so that TUG funda will very likely not be required Bob s new position in favor of institutional support is not affected by this change a fact he has confirmed in a telephone conversation 4 REPORT ON THE IMPLEMENTORS WORKSHOP STANFORD 14 15 MAY 1981 Barry C W Doherty At the Implementors Workshop in May 92 people were registered a complete list follows The goal was to draw together both those knowledge able about and those in various stag
19. DEF LOCK Hi7WATER FIG 4 MACRO VIRTUALIZATION ANOTHER VM SYSTEM OUTPUT ROUTINE 1 OUTPUT ROUTINE 2 ALIGNMENT TEMPLATE 1 9 4096 8192 BINARY POINT FIG b GLUE FORMAT 31 16 15 LINE BREAKING WIDTHS MILS F16 7 DATA FORMATS TUGboat Volume 2 2 5 k k amp Output Devices E k g OUTPUT DEVICE NEWS FLASB APS 5 AND LINOTRON 202 David Fuchs A quick note just before the deadline During this summer I will be working on Pascal interfaces for the Autologic APS 5 and the compatible micro 5 and the Mergenthaler Linotron 202 The first incar nation of each interface will allow TEX to use only fonts supplied by the manufacturer There is some reason to hope however that I will also be able to get enough information about font encoding on these machines so that I can write METAFONT output modules for each one This would allow the CM font family as well as any other METAFONTed fonts to be used with these machines I am particularly enthusiastic about the APS micro 5 the native lan guage of the machine looked good the documenta tion seemed complete and the people at Autologic were quite helpful They even said that they would consider releasing their font encoding description on a non disclosure basis if it helps us sell typesetters For more details on the status of either of these in terfaces please contact me
20. David Fuchs that prints the contents of a DVI file In summary we really can see some light at the end of the tunnel Michael J Frisch DECSystem 10 20 IMPLEMENTATION WORKSHOP Phil Sherrod An implementation workshop for persons inter ested in installing TEX on DECSystem 10 or 20 will be held at Vanderbilt University in Nashville Tennessee on September 10 and 11 The course will TUGboat Volume 2 2 discuss the details of connecting TEX output devices to DEC 10s and 20s as well as software issues in volved in installing Conference attendees will have access to a system running TEX For details contact Phil Sherrod Vanderbilt University Box 1577 Station B Nashville TN 37235 615 322 2951 AMS SITE REPORT Barbara Beeton Since our last report TUGboat Vol 2 No 1 page 48 an integrated spooling system has been developed and installed to control output to both the Varian and the Florida Data This spooler will print straight text ASCH files using the rather primitive fonts supplied with the output devices as well as TEX output Output defaults to the local device using device appropriate fonts Varian in Providence Florida Data in Ann Arbor but may be overridden output page positioning may be ad justed either interactively or from a predefined op tion 0 file allowing each column of a multi column page to be treated by TEX as a separate page to
21. Euler family can be made generally avail able to other printing houses 4 ko kx Ronald Whitney American Mathematical Society was mentioned in Bill LeVeque s article see page 39 the AMS has been using METAFONT to generate a set of small highly condensed fonts to be used in the Combined Membership List of the Society and two other mathematical organizations We started by creating a 6 point sans serif Computer Modern font and narrowed and thinned it until parameter twiddling no longer yielded significant improvements Since proofmode was not yet avail able here and pens at Varian resolution were 1 2 pixels wide more proofing than usual was done with the Alphatype Inter letter spacing seemed to provide the most difficulty As can be expected when a font family is pushed very far in a direction in which it is not designed to go it became necessary for our purposes to bifur eate some parameters serif correction unit width one pen size and to modify some drawing routines The result is our first approximation to what we have called the telephone book fonts The TUG membership lists in this issue have been printed in TPHON and welcome comments and criticisms amp bout its readability k PROOFMODE AND MAGNIFICATION Barry C W Doherty and Ronald F Whitney American Mathematical Society A large part of the effort at the AMS
22. FinalEnd in TEX Again this was straightforward 3 Some new features were added to the lo cal compiler by Andy Hisgen to support ASCII files and False starts FILE OF ASCII does the expected thing here except the conventions for RESETing the terminal are somewhat different FalseStart is like the MACLISP SUSPEND opera tion If a Pascal program calls FalseStart then execution is suspended and the program may be SAVEd When this core image is STARTed up ex ecution will resume at the FalseStart call I added such a call to our copy of TEX PAS just before the cali to InitSysDep 4 The installation documentation was reason able though it could be a bit more detailed in certain areas Examples expected problems the symptoms of various bugs e g not reading the STRINI file some remarks on the control structure of TEX 5 Testing here has been a bit skimpy since I can t easily get hardcopy output 6 Some hacking still remains I haven t touched AppendtoName yet but I expect no problems here Andy Hisgen suggests changing the procedure er ror so that ordinary letters are used instead of CR and LF Thus the help message becomes something like or to continue f or to flash error messages 1 or 9 to dismiss the next 1 to 9 tokens of input I to insert something x or X to quit instead of Type cr to continue lt 11 gt to flash error messages 1 or or 9 to dismi
23. This is so that accumulated errors will be insignificant even in the worst imaginable case a box many feet long The current position on the page is moved about by the commands W0 W2 W3 W4 X2 0 Y2 Y4 ZO 22 23 and Z4 The vertical coo dinate is changed by Y and Z commands while the horizontal coordinate is changed by W and X commands The value of horizontal coordinate can also change as a side effect of setting character or rule VERTCHAR and VERTRULE commands the current position on the page moves right the natural width of the character or rule set The POP command may also change current position on the page So whoever or whatever reads a DVI file might have three variables F and V to keep track of the current font and the current position on the page Four more variables are also called for w amount x amount y amount and z amount These variables hold not locations but distances in rsu s The amount variables are used in DVI files to move the current position on the page around The commands X0 and W0 add x amount and w amount to horizontal coordinate respectively while YO and Z0 add y amount or s amount to vertical coordinate respectively There are also number of commands 13 that change the value of w amount x amount y amount or s amount W2 W3 W4 X2 Y2 Y3 Y4 Z2 Z3 and Z4 these commands also change horisontal coordinate or vertical coord
24. ascii character or a question mark if the charac ter is non printing glue as a space and hyphena tion nodes as dashes This gives you an idea of whether TEX knows how to hyphenate a word at the beginning or end of the line and you can insert dis cretionary hyphens as needed However if the word that is giving trouble occurs on the previous or next line you must consult TEX s hyphenation rules to find whether it knows how to hyphenate the word Thus I have found myself hyphenating words at ran dom which is a distinct inconvenience in itself and often find that there are still overfull boxes In addition to poor diagnostics we had trouble in getting TEX to set the narrow columns at all without overfull boxes We tried adjusting the jpar parameter to have TEX look at more pos sibilities and it still gave overfull boxes Manual hyphenation did not work We eventually had to in crease the spaceskip or word spacing parameter to allow very large space between words and create typographically bad output In some cases even this did not work and a linebreak command had to be used At we charge customers as closely as pos sible for the resources they use and we have found TEX to be very expensive in both memory and CPU time This is not really a surprise to anyone of TUGboat Volume 2 2 course But it means that most users cannot afford to run a document through four or five times to find
25. collecting statistics on the effectiveness of all possible patterns and then using some heuristics to choose good set of patterns Preliminary experiments with this approach indi cate that it will be very effective For example a set of about 3300 hyphenating and 2700 inhibiting patterns gets 85 of the hyphens with 2 error TUGboat Volume 2 No 2 21 7 on SMALL MACHINES Kent S Harris and Robert M McClure Unidot Inc Sunnyvale CA 1 ABSTRACT As small computers become more affordable the demand increasingly sophisticated software grows Unfortunately much of this software is large and does not easily fit on small computers Emu lation through interpretive techniques typically yields unacceptable performance Reengineering of existing software is frequently desirable therefore for reasons of beth cost and compatibility The art of compressing large mainframe developed programs into small address apace machines has become very important One such effort is the reengineering of the TEX typesetting sys tem Since this paper really deals with putting large software systems on small computers and is not a paper on itself familiarity with is assumed 2 INTRODUCTION is a recent creation by Donald Knuth at Stanford University It is a system for typesetting beautiful books especially books that contain a lot of mathematics an area in which the cost effectiveness of computer
26. dvi files will be much easier to solve than the problem of getting the files printed because of the variety of printers used and the secrecy about their inner workings Perhaps this should be the next major problem that TUG could make a sys tematic attack on Of course TEX is already up and being exten sively used at some places and more and more macro packages are being produced to get TEX to do just about everything except shine your shoes and write the papers for you At the present stage there are clearly still many tricks to be learned as Don said we are just beginning to scratch the tip of the iceberg Even if macro package performs some function that isn t of particular importance to another macro writer it may contain some tricks that will be useful Perhaps we should encourage more people to send in special tricks or emphasize such tricks in their macro packages eventually standard library of tricks could be compiled Hours of pestering Don have produced some basic tricks documented in the article Macro Madness see page 50 that may help people to make macro facility work more like the programming lan guage that many have wished for As this last paragraph has indicated my own par ticular interest in using TEX is to get it to typeset anything a mathematician would want with minimal understanding on the part of the typist Obviously the interests of other users and implementors
27. for the collection and distribution of these macro packages machine readable form probably on magnetic tape details will be published as soon as they are known Fiche will conform to the following specifications negative image white characters on black 105mm x 148mm 24 to 1 reduction ratio containing 98 frames per fiche ERRATUM NOFILL PROGRAM Patrick Milligan BNR Ine There was one subtle error in the program listing of both the SAIL and Pascal versions of the NOFILL program that appeared in TUGboat Vol 2 No 1 In both programs the definitions of macros V and V were reversed see pages 90 and 96 As printed the definitions are correct but the program source was incorrect Since the program source was run through NOFILL for publication the incorrect definitions became correct but all other uses of 7 acute accent and grave accent were incorrect Also there was some confusion about the table of contents entry on page 136 entitled NOFILL Program with Pascal Source When the two pro grams were submitted to TUGboat it was not clear if the SAIL version would be printed or the Pascal version or both introduction to the SAIL ver sion was appropriate to both versions but no intro duction was prepared for the Pascal code 10 READFONTINFO A A AG M 5 I meee TUGboat Volume 2 No 1 _THX PASCAL SYSDEP Procedure Specifications
28. gotten them all because we haven t had subscript out of range errors but it did mean that all the memory reduction was from the higher end of mem which is probably not optimal We occasionally run into TEX capacity exceeded memsize 25000 I didn t try to alter the other memory parameters like varsize because there were so many instances of varsize i and such that would have been affected We ran into another interesting problem on a UNIVAC a person typing at a terminal can type eof and his terminal input is considered to have reached an end of file This concept doesn t exist on most systems so it wasn t considered in TEX Basically if a person types Geof I artificially return end to TEX but this doesn t always work TUGboat Volume 2 2 I need to do more work on this If this affects other sites this is something you might want to look into CMUA PDP 10 HAMBURG PASCAL BILL SCHERLIS 1 Some changes in the code were required in or der for compilation to succeed here In particular the local compiler uses different conventions for PACK and UNPACK has different switches and does not want a PROGRAM statement Also a main program body is not required in a file for separately compiled procedures These changes were all fairly minor 2 The compiler here is not friendly to inter procedural GOTOs so these were eliminated by add ing a new WrapUp procedure See the labels endOfTEX and
29. has been the development of a library of fonts adequate for our typesetting requirements and for this we have been making extensive use of the SAIL version of METAFONT So far this has taken four direc tions 1 creation of fonts in larger point sizes e g 14 and 18 point 2 filling in holes e g 8 point text italic 3 production of a complete set of fonts for the Florida Data and 4 creation of a telephone book font cf the article by Ron Whitney in this issue p 40 TUGboat Volume 2 No 2 All but 3 have necessitated changing the parameters in the MF files but only with the last have really significant changes had to be introduced to compensate for the changes of parameters sub jected to excessive shrinkage While METAFONT has been able to generate the Florida Data fonts it has not been able to cope with the rounding prob lems made more visible by the low resolution of the output some improvement might have been had by modifying the MF files at the expense of in consistency of fonts across devices leaving as the only practical alternative the manipulation of the raster pattern via an intermediary file Changing the font description would also violate Knuth s goal of designing to the highest resolution device in our the Alphatype CRS This development has been able to proceed without the availability of proofmode although 4 concerns the creation of new font and would have been
30. maintain MS but automatic changes caused by things like over will go undetected In such cases the user must define MS himself if it is going to be tested sample application here is definition of bf select bold face which behaves the same as normal in non math mode and selects a font of the right size in math mode In the latter case it acts only on the following character control sequence or group Let s suppose that A B and C are bold face fonts of the required sizes det bf itmmode gdef Fnt 1 hbox it We T V A 1 else it us s B 1 else c 1 elae gdet Fat A Faot In our second application we ll define a macro for raising a portion of text If you type lift text by dimen then text is put in hbox and raised by an amount dimen text will appear in the current style except that display style and text style are not distinguished def 1ift 1 by 2 raise 2 hbox ifamode MfWS T 1 else ar us 8 scriptstyle i else TUGboat Volume 2 2 2 Groupless 17 good source of inscrutable bugs involves the way that TEX handles conditionals like 11 ifpos ifvmode etc Let s suppose that we want to select font A if the macro format has the value 1 and font B otherwise The obvious method is it formati A else B but this doesn t work The reason is that the text produced is not A or B but
31. not have to undertake font management and spooling tasks BASIC SYSTEM Four fonts eSorter collator ePage orientation flexibility eRS 232C interface By selecting a suitable combination of features and an interfacing arrangement this versatile printing system can be adapted to wide range of tasks for word processing business data processing communications in plant printing and graphics IMAGEN 12769 Dianne Drive Los Altos Hills CA 94022 415 949 1580 27 28 Site Reports ko k UNDER THE NORTH STAR Part I coordinator for the Cyber implementa tion of 1 have agreed to answer questions from interested Cyber sites about our progress with in Pascal and with our implementation of drivers for the output devices to which we have access When we have succeeded in our implementation I will make available for distribution for a charge amp tape with the following information 1 source as modified to com pile successfully on Cyber under the ver sion of Pascal maintained at the University of Minnesota 2 A file containing the Computer Modern font set 3 Our device driver s 4 Relocatable binary of these files I will also send a copy on microfiche of the Pascal compilation and cross reference listing I will provide sympathy understanding and hand holding as you put up on your system I have also agreed to act
32. other words macros can directly or indirectly call themselves Of course we are not given this little gem of information because the knowledge would be almost useless Nevertheless there is a little gap between almost useless and completely useless and this section is devoted to exploring it Three applications of recursion we will consider are i loop structures ii counter arithmetic and iii macros accepting variable numbers of arguments i Loops are quite easy to create in TEX as long one respects IEX s finite capacity In order to make loops which can be repeated a large number TUGboat Volume 2 2 of times the recursive call must the very last thing in the expansion and in particular it must not be in a group TEX won t nest groups to an indefinite depth The last requirement means that the recursive call can t be part of the result text of a conditional see Section 2 Here s some examples Met savecount 1 2 ifpos i xdef 2 count 1 else setcount 1 count 1 xdef 2 count 1 setcount 1 count 1 Mgdef Wtemp 1 2 2 Wloop i 82 def Wloop 1 2 ifpos 1 else gdef Wtemp 1 2 wtenp 1 2 def while 1 2 endwhile Wloop 1 2 gdef Wtemp 1 2 2 Wloop 1 2 def repeat 1 times 2 endrepeat savecount9 Rtemp setcount9 1 while9 2 advcount9by 1 endwhile setcount9 Rtemp savecount digit control sequence saves the value of a counter in a control sequence whil
33. presented in the next issue Readers with working systems are encouraged to attempt solutions to these problems in order to better appreciate the problems and their solutions Lynne A Price Problem no 1 Type 12pt noindent hide Allan Temko vskip 2pt noindent Architecture Critic To get Allan Temko Architecture Critic TUGboat Volume 2 2 55 Problem 2 fancy Senator and Mrs Stanford had reserved to themselves control of the University s affairs during their lifetimes including the parceling out of 77811 the money that could be wisely used Mrs Stanford had remained in her husband s shadow on opening day she could not bring herself to deliver the short speech she had written out But following the death of the Senator she at age 65 took on full responsibility for the University with unsuspected strength To get enator and Mrs Stanford had reserved to themselves control of the University s affairs during their lifetimes including the parceling out of all the money that could be wisely used Mrs Stanford had remained in her husband s shadow on opening day she could not bring her self to deliver the short speech she had written out But following the death of the Senator she at age 65 took on full responsibility for the University with unsuspected strength Problem no 3 Type hsize 25em noindent This is case where the name and address fi
34. records which in turn may be translated by DVIPDP one can see by the above description us ing TEX at NIH is complex The four steps TEX DVIPDP CLINK transfer to the 11 and TEX 11 will need to be compressed before TEX can come up in production environment This necessarily in volves getting our own small computer and printer to hook up to the DECsystem 10 for usage with TEX NIH is community of scientists and researchers who produce technical papers in abundance The potential for use of TEX at NIH could be very high However TEX is not an easy system to learn We are hoping that the 5 macros will make us ing TEX a little easier and we are looking into the possibility of writing some type of TEX preprocessor x ks AN IMPLEMENTATION REPORT FOR THE UNIVAC 1100 Bill Kelly University of Wisconsin Madison We have running on the Univac 1100 using University of Wisconsin Pascal This is a report on some of the problems we encountered in implement TUGboat Volume 2 No 2 ing TEX The difficulties were of several kinds A major difficulty is in the differing syntax of exter nal compilation between various Pascal compilers a feature not defined in the original Pascal definition Differences in I O on the host machines and in the ways the compilers handle I O were another prob lem Memory limitation problems were encountered and these were made worse by the fact that w
35. remember that tricka is really tricka so it can t appear in user s file We have to consider two possibilities for in order to determine the result of this construction Suppose first that is not empty Then argument 1 for empty will be the first token or group of and argument 2 will be whatever remains if anything Hence empty tricka tricka trickb gt trickc tricka trickb false trickb gdef ans F But suppose that is empty so that we have empty tricka tricka trickb Note that argument 1 for empty must be non empty since it is not followed by a token in the 51 definition of empty So in the present case ar gument 1 for empty will be the first tricka Consequently the second tricks will play the role of the token tricka in the definition of empty and argument 2 will be empty Thus empty tricka tricka trickb tricke trickb gdef ans T In other words empty tricka tricka trickb defines ans to be Tif and F otherwise is empty There would appear to be one exception to this rule if is a blank space or a sequence of blank spaces then ans will still be defined to be T since spaces after the control sequence empty are ignored But in practice will always be an argument from some other macro and in this case the exception does not arise Suppose for example that we define def try 1 empty 1 tricka tricka trick
36. tape 600 foot in BCK RST format that s an MCR utility for backup restore which works well for binary files We can write 800 or 1600 b in tapes hearing noth ing we ll send 1600 Fees fifty dollars will get you tape and ship ment via UPS To minimize our overhead please don t send a tape and do send a check or negotiable securities so we don t have to deal with purchase or ders and billing I d prefer small unmarked bills If you or your friend are in the truly needy just let me know Maintenance We ll be working on bugs and con venience improvements for the perpetual future will be pleased to hear commente suggestions gripes bugs no promises of any response except through the TUGboat and Monte s newsletters k k Fonts x Barry W Doherty Getting TEX up and running is only the first step in producing output One needs also an output device with an appropriate driver and fonts Transportable TEX output requires compa tible fonts Those sites which have similiar output devices Varians etc and are willing to use the Stan ford Computer Modern family have this problem solved There are many however who feel limited by the CM family as available from Stanford either for reasons of design completeness or merely preference for more traditional fonts Helvetica for instance TUGboat Volu
37. tell where the characters on the page go followed by an EOP command Each EOP command is im mediately followed by another BOP command or by the PST command which means that there are no more pages in the file and the remaining bytes in the DVI file are the postamble Remember that TEX really doesn t have an official knowledge of page numbers although it does print the value of count0 on your terminal as it outputs each page on the as sumption that some meaningful number is there so the only thing that can be said about the order ing of pages in a DVI file is The order in which pages come in a DVI file is the same order in which TEX constructed them which is the same order in which the user specified them Any blank or nonexistent page from a TEX job might not be in the DVI file at all If we consider the page number to be the value of NXcountO then the page following page number 34 DVI file might well be page number 5 Some parameters of DVI commands are pointers pointer is simply byte number as discussed pointer itself is 4 bytes long For example BOP command s last parameter 4 is the BOP s previous page pointer This parameter is the num ber of the byte in which the previous page s BOP command begins In particular the second page s BOP command s previous page pointer parameter p 4 is always zero since the first page s BOP is always in byte zero a DVI fil
38. the ASCH character set and those that do may limit or prohibit the use of ASCII control charac ters There are standard translations be tween ASCI and EBCDIC graphic charac ters but no such translations exist for con trol characters TEX can usually avoid the use of control characters but as we have seen in recent TUGboat macro packages and in the TEX manual itself it is tempting to use TUGboat Volume 2 2 the extended ASCII character sets in use at Stanford MIT and CMU if they are available In addition control sequences for nega tive conditional thin space lt and conditional thin space gt must be entered using control characters TEX makes some assumptions about the under lying structure of text files In particular it is assumed that a file is organized as a long string of characters which is divided into lines by end of line characters and into pages by form feeds On some systems the structure of text files is either fixed length card image records padded with blanks and possibly with sequence numbers in columns 73 80 or vari able length records rounded to computer word boundaries and padded with blanks or some other filler In most cases it is not impor tant to know where the placement of the end of line is or whether the trailing blanks on line are real or supplied by the system However if the meaning of blanks or end of line characters is changed throu
39. the right discretionary hyphens needed to avoid an overfull box Nor is this user friendly There may be no easy way to solve this justification problem It may be an unavoidable side effect of the justification algorithm used One possibility would be having a paragraph debugger that would help to diagnose the problem in a given paragraph e g point out a large word that can not be hyphenated that is at the root of the prob lem This would be a difficult program to write An alternate hyphenation routine would be another possibility Allowing to violate its hyphena tion rules if necessary to justify a paragraph mean time issuing a warning would at least focus on a specific problem i e a questionable hyphenation rather than leaving the user guessing Usefulness of macro sharing TEX in itself seems to provide a standardized lan guage for typesetting However a non computer programmer will find this language difficult to use and will not have a ready facility for writing his own macros users need to do much more macro sharing as was done in TUGboat Vol 2 No 1 Good macros put the typesetting capability in the reach of most users There are still problems when an error occurs through improper use of a macro The diag nosties are geared toward one with a knowledge of programming One aid in this would be to provide amp way for macro to send a message directly to the terminal This was shown in TUGboat
40. typesetting over manual methods is obvious The equation VF Xm joo e vivid example That is especially popular today compared to other computer typesetting sys tems is due primarily to TEX s level of sophistication and ease of use has recently been endorsed by the American Mathmatical Society for submission of machine readable input a trond that is likely to grow quickly in the years ahead _ The goal of this project was to implement TEX on a small preferably desktop size computer We 1 The X in TEX is actually Greek chi and therefore is pronounced tho same the first syllable in technology The name TEX is a registered trademark of the American Mathemati Society CH1626 1 81 0000 0380 00 75 1981 IEEE 380 also wanted to use a commercially available operat ing system that would provide additional typesetting tools such as editors and other text processing facilities Since UNIX had demonstrated superior text manage ment functions over other operating systems and ap peared as if it were becoming a de facto standard for small machine operating systems it was chosen as basis for this implementation UNIX was in the process of being ported to several 16 bit architectures by various manufacturers while already enjoying rather widespread usage in the DEC PDP 11 Another issue was that of implementation lan guage The C language seemed to meet the require ments best f
41. 131 p 4 14 a 4 m 4 4 w 4 Fontdef Fontdef Fontdef 1 4 q 4 1 1 223 The postamble starts here See below for the full explanation of the parameters of the postamble Commands 218 255 are currently undefined and will not be output by TEX The PST command which is always the last com mand in a DVI file is somewhat special The parameter p is pointer to the of the final page in the DVI file The parameters n and d are the numerator and denominator of a fraction by which all the dimensions in the DVI file should be multiplied by to get rsu s TEX always outputs 1 for each of these values they are included in DVI format to allow other text systems to con veniently output DVI files The parameter m is the overall magnification requested by par12 in the job 12 is unitless and is 1000 times the desired magnification Next come h and which are the height of the tallest page and the width of the widest both in rsu s Next in the postamble come the font definitions one for each font used in the job i e each FONT and FON TNUM command in a DVI file must refer to amp font number that has a font definition The format of a font definition can be considered to be fnum 4 f chk 4 fmag 4 fnamlen 1 fnam fnamlen The font number is held in fnum The font checksum from the font s TFM file is in fchk The parameter fmag holds the font magnification 1000 times the at size o
42. 2940 Submissions to TUGboat are for the most part reproduced with minimal editing Any questions regarding the content or accuracy of particular items should be directed to the authors TUGboat Volume 2 No 2 OFFICIAL ANNOUNCEMENTS 1982 Membership Dues 1982 dues for individual members of TUG will be 15 Membership privileges will include all issues of TUGboat published during the membership calendar year new members and other persons inquiring about TUG will be sent TUGboat Vol 1 No 1 but 1981 issues will be sent only to persons paying the 1981 dues of 10 Beginning in 1982 foreign members will be able on payment of a supplementary fee of 12 per subscription to have TUGboat air mailed to them TUGboat Schedule Volumes of TUGboat are numbered on a calendar year basis Volume 1 appeared in 1980 Volume 2 corresponds to 1981 and 1982 will bring Volume 3 Volume 1 consisted only of issue No 1 dated October Three issues are planned for Volume 2 No 1 appeared in February and No 3 is planned for November No schedule has been determined yet for 1982 The deadline for submitting items for Vol 2 No 3 is October 1 1981 Contributions on magnetic tape or in manuscript form are encouraged editorial addresses are gwen at the bottom of page 2 and a form containing instructions for submitting items on tape is bound into the back of this issue It has been necessary to reprint back issues of TUGboat to fulfill the
43. CHOOL UNIVAC SERIES 90 PASCAL 8000 GEORGE OTTO Pascal 8000 1 2 does not accept numbers like 100B or 400000B These numbers must be changed to the appropriate integer or real form Pascal 8000 1 2 does not support EXTERN proce dures and functions because of the internal loader Pascal 8000 1 2 uses ELSE for the default case of CASE statements Not OTHERS like Hamburg Pascal Tab characters not allowed in source 18 Our Pascal must uniquely distinguish between all identifiers in the first 8 characters Longer identifiers can be used but only the first 8 characters of them are significant At the moment we are having trouble writing tape from our EBCDIC machine to be read by Wharton s ASCII machine to be sent to you over the net No standard MAX and MIN functions 8000 1 2 has problem recognizing 10000000000 0 as amp real The fix is to use 1 0E10 Pascal 8000 1 2 stores elements defined 0 255 and 0 65535 in 32 bit words Records are as signed to length of the longest freevariant possible Therefore the memory structures of TEX will not work as is U OF MINNESOTA CD CYBER MIKE FRISCH Everything must be uppercase Can t pack memoryword properly this is bad Had to replace OTHERS by OTHERWISE JET PROPULSION LAB UNIVAC 1100 81 U WISCONSIN PASCAL CHARLES LAWSON This compiler employs environment modules CD made one containing outer block TYPE and EXTERNAL procedure declara
44. Owing to the very high cost of mathematical printing it has been necessary to limit the alterations in this edition in order that as much as possible of the earlier book may be reproduced by photographic processes To save expense I have refrained from small alterations which would have been made were I starting afresh to improve the English amp c being content to let what was printed stand provided it was reasonably clear and not actually wrong Harold Hilton Plane Algebraic Curves Oxford University Press 1931 THE USERS GROUP NEWSLETTER EDITOR ROBERT WELLAND VOLUME 2 NUMBER 2 JULY 1981 PROVIDENCE RHODE ISLAND USA BALL George Computing Center Washington State University Pulman WA 99164 509 335 6611 BEEBE Nelson Department of Physics 201 North Physics Buiding University of Utah SaR Lake City UT 84112 801 561 5254 BEETON Berbera American Mathematical Society Box 6248 Providence 02940 401 272 9500 415 327 0483 DONERTY Berry American Mathematical Society Box 6248 Providence Rt 02940 401 272 9500 FRISCH Micheal J University Computing Center 208 Union St SE University of Minnesota is 55455 612 313 4599 FUCHS David Department of Computer Sciance Stanford University Stanford CA 94305 415 497 1646 GOUCHER Reymond E American Mathematical Sonety 6248 Providence R3 02940 401 272 9500 HARRIS Kent
45. The limit on the size of wxy subscripts was lifted e Use of lt gt gt and lt for SAIL s one character relational symbols e Limitations on the size of the raster currently 300 x 300 e Output routines adapted to drive Anadex printers The conversion went fairly smoothly the major problems being the lack of SAIL s nice strings and the difficulty in determining exactly what field in a word the SAIL version was using and what implicit initializations were being performed by shifts used to access and load fields Experience with METAFONT In general working with METAFONT is pleasure The interpretive nature of its execution allows number of powerful tracing mechanisms which are very useful In designing an alphabet the most difficult problems are deciding on the basic approach to be taken in specifying characters and writing the base routines which support such an ap proach Once this is established adding characters is fairly straightforward Most of the problems encountered in writing METAFONT programs arise from the complexities of the characters themselves not METAFONT but there are several features which users of other pro gramming languages will occasionally miss e Loops e Combinations of Boolean expressions e full complement of intrinsic functions e Ability to write functions OCLC Online Computer Library Center is a not for profit computer library service and research organization based in Dubl
46. Vol 2 No 1 as an extension added by site it seems to be useful enough feature to incorporate into the stan dard TEX k AVAILABILITY OF i OREGON SOFTWARE IMPLEMENTATION OF TgX FOR THE VAX VMS Monte C Nichols This VAX TEX site report consists of an ab breyiated version of memo recently sent to all VAX TEX users The information contained here should allow anyone who obtains the VMS version from Oregon Software to implement same on their VAX VMS system only output device sup ported at this time is a Versatec 1200 A or 80 printer having a DMA interface to the VAX preliminary VAX VMS implementation of TEX with auxiliary programs to support Versatec printer is now available from Oregon Software 33 Further progress on a UNIX TEX at Brown awaits the arrival of the U of Washington compiler The accompanying article by Barry Smith explains how you can obtain a copy of the VMS version You should understand that in spite of the efforts made by Oregon Software this is not the final version in fact Stanford has just released a new version of TEX Barry has overcome numerous bugs in DEC Pascal to get us to our present state but more remains to be done You can help in this effort by carefully documenting any bugs that you encounter after making sure insofar as possible that they are bugs and not errors you are making during your TEX learning process We have en
47. Zabala Pascal related issues Concentration on the characteristics and suitability of various popular Pascal compilers with suggestions on what to look for in a compiler and how to cope with the compiler one has See this issue p 16 11 15 12 00 Ignacio Zabala The system depen dent module of TEX in Pascal The Pascal elements of and their implications See the artecles by Lawson Zabala and Diaz TUGboat Vol 2 No 2 pp 20 32 1 15 2 00 David Fuchs Different output formats conversion issues Largely a discussion of DVI file format See this issue p 12 2 00 2 45 Luis Trabb Pardo From DVI to paper General discussion of translator driver pro grams from DVI to something a specific device understands and the role of spoolers servers in scheduling and queueing features and characteris tics downloading of fonts memory requirements efficiency 3 00 3 30 Frank Liang Hyphenation in Discussion of the algorithm used in TEX and com p rison with other widely used algorithms See this issue 19 3 30 4 00 Michael Plass Lines paragraphs pages Discussion of how TEX functions in this con text See report by Donald E Knuth and Michael F Plass Breaking paragraphs into lines Stanford CSD report CSD CS 80 828 A panel discussion had been scheduled to begin at 4 00 talks were running longer than planned as a result of discussions following most Instead Don Knuth spen
48. a useful version should be finished this summer While it is anticipated that Chel will be made available outside OCLC no distribution mechanism has been estab lished TUGboat Volume 2 2 The letter b File created by METAFONT 06 05 81 08 38 29 AM The ietter b a ato eee 444 917225229999 DEM Te 4 Figure 1 Lowercase b Proofmode at OCLC 37 TUGboat Volume 2 2 38 Chel font at OCLC Figure 2 Variations on the letter b TUGboat Volume 2 No 2 FONT DEVELOPMENT AT THE AMERICAN MATHEMATICAL SOCIETY William J LeVeque As Don Knuth has declared his primary pur pose in developing METAFONT and TEX was to be able to produce new editions of his book that would look like the originals So it was reasonable that he should focus his attention while designing par ticular alphabets and character sets in the Computer Modern family on those fonts needed for composing ordinary text and mathematics Roughly speaking these were Computer Modern Roman Bold Face Slant Sans Serif and math mode Italics Greek symbols and script in 5 6 7 8 9 and 10 point sizes and Upright Italic and Typewriter in 9 and 10 point sizes When the American Mathematical Society AMS decided to continue the development of Knuth s sys tem so that it could be used for the production of a wide variety of Societ
49. ager can install them with these privileges so they can be used by any unprivileged user Although you could start your first TEXperience by using TEX interactively or with a file built us 34 ing your system editor using Knuth s and METAFONT as guide it is suggested that you might want your first attempt to be with the file TYPO TEX To run this example from the TEX directory without using the startup command file provided enter R Wait for the prompt this may take up to 20 seconds input typo Note lowercase Watch the TEX comments wait for system R LVSPOOL LHSPOOL the system should handle the rest The mailing list for future informational memos regarding VAX TEX will be the list of DUES PAYING TUG members who indicate a VAX interest make sure you join FOR VAX VMS Barry Smith Oregon Software Well it works TEX for the VAX running VMS is alive available and in production use Production use is defined by example we ve just finished 192 page manual for our optimizing PDP 11 Pascal compiler that is entirely typeset by TEX including charts and diagrams It s not yet perfect nor something a naive user should be expected to enjoy list the major deficiencies e There are still annoying bugs such as the flaming crash that occurs when one tries to in sert a footnote Most of these seem to be due to bugs in Digital s VAX Pascal which just ma
50. alized these local quirks To make it easier for future sites to implement TEX procedures like SkipTrailingStuff to ignore nulls after a line end and functions like EndOfLine which would check for PDP 10 line end characters but note that on non PDP machines a function like EOLN might be used The fact that expects input files to be page oriented is a minor nuisance Apparently files on the PDP 10 are made up of pages delimited by a form feed character This is not the case on our machine and on many others This causes minor problems in two respects one is that if a file containing a form feed character on our system were read by it would be interpreted in manner inconsistent with our operating system The other is that page numbers appear along with line numbers in the error messages This tends to be very confusing since there is no page number in the file structure but there are page numbers in the formatted output so Univac user would assume the page number referred to the output pages We circumvented this by changing the error message from 0 1 50 to Line 50 This could be built into TEX by providing a Pascal constant pageOriented which the TEX implementor could set to true or false Implementing is difficult because the pro gram is written in a macro language which is converted by the UNDOC program into Unfortunately the UNDOC program is not presently
51. anford has attempted to solve this problem for the distribution of TEX in Pascal macros and fonts The current organization of site coordinators has solved the problem of distribution between sites us ing similar computers through the use of common operating system dependent tape formats However the problem of general machine independent tape interchange between TEX users who use different computers has not been completely solved is im portant that standards for tape interchange be es tablished and portable tools developed to support these standards One potential area which TUG should explore is the sale of machine readable macros and programs submitted to TUGboat Having one distribution center for these contributions would be preferable to contacting the author s of a particular program or macro package Receiving one tape from TUG would be easier than requesting tapes from mul tiple sources and would be much easier than typ ing in part or all of a long macro package pro gram In addition TUG would have another source of revenue This sort of scheme has worked well for the DECUS Library a part of the DEC Users TUGboat Volume 2 No 2 Society and for Addison Wesley s distribution of Ratfor source for the programs in Kernighan and Plauger s Software Tools In conclusion it is clear that what you get out of TUG depends on what you are willing to put into it Without member contributions there would be no TUG
52. as go between for other matters which may come up from time to time However I don t intend to act as a roadblock Bob Welland editor of TUGboat would be delighted to receive articles and letters to the editor directly from you My address is in the front of TUGboat my phone number is 612 373 4599 This number is always answered and I am reasonably good about returning calls In Part II of this report J Frisch our TEX im plementor describes our current progress However please note that calls should be directed to me not to him Thea D Hodge Part We recently got proper device independent DVI file output from our Pascal version of This is version we received from Stanford in December 1980 While we are pleased with this success it is only a beginning Several temporary changes had to be made to get TEX to work because our conversion is incom plete Our font information file had only one font which was converted to 60 bit Cyber words from 36 bit DEC 10 words containing mixed integer and TUGboat Volume 2 No 2 floating point The input text was very simple and the output was only a set of numbers the contents of TEX s DVI file We have no device driver programs and our next major task is to write one or more Potential devices at our site are Varian 200 dot inch plotter a Linotron 202 typesetter plus amp Xerox 9700 page printer at a local service bureau The 9700 would be inten
53. at this address David Fuchs Computer Science Dept Stanford University Stanford CA 94305 I am also interested in hearing any news of anyone succeeding or failing to interface TEX to any device ks k amp Summary of Computer Equipment and Output Devices The following computer amp rchitecture groups and output devices have been specified by TUG mem bers Computers marked with are known actually to have a version of TEX installed and capable of producing DVI files output devices so marked have actually produced output copy from DVI files No distinction is made between systems capable of production and those still operating only on a test basis single indicates that work is in progress The separately bound membership list contains sublistings of member names by the device types given below is not yet any cross reference mechanism to indicate which output devices are connected to which computers for details see the 25 individual member listings and the Site Reports column which appears in every issue of TUGboat COMPUTERS Amdahl Ithaca Intersystems Apple LSI 11 Burroughs microcomputers CDC Motorola 68000 CDC 6400 Multics CDC 6600 eeNord CDC 6700 North Star Horison CDC 7600 esOnyx Cyber PDP 11 Cray PDP 8 DEC PERQ 10 Pascal microengine 20 Perkin Elmer DEC WPS 8 Prime DG Raytheon 703 DG Eclipse 8 100 DG Nova Siemens DG
54. ating my worries about the effectiveness of TUG The Steering Committee is extremely reluc iant to adopt any formal structure or bylaws We certainly want to avoid unnecessary regulation and such looseness is fine as long as it does prevent the committee from functioning We do want to impose certain constraints I believe the Steering 57 Committee did decide for example that while each of its members is free to define his own r le site coor dinators should not discourage relevant telephone calls The committee members are very aware that opinions differ and are reluctant to take action that might impose their views on the group as a whole I fear that this admirable attitude in con junction with an informal structure may result in an ineffective Users Group As a case in point Sam Whidden mentioned in May that the Steering Committee had decided against assigning the main tenance and distribution of TEX to a software house There was considerable discussion of this point in January Bob Morris eloquently argued about the dangers to university users of such an approach I was not aware however that Bob had succeeded in convincing the committee as a whole I had supposed that the finance committee would have prepared alternate proposals before this last meet ing that there would have been more discussion and that a final decision would have been based on a vote Certainly we cannot continue to abandon proposals sim
55. ating them into editable form and repacking them More such pro grams are needed and they should be as transport able as This could also help with the prob lem of disseminating font information avoiding bi nary data entirely There is also a need for device independent fonts at least to the extent of uniform font metric files and a reasonable representation for proofing so that a TEX file from an institution that had say Helvetica could have that file printed else where Another role for the committee might be as source for model RFP s to aid those institutions which must submit such documents to vendors this would simplify the task of acquiring printers and fonts We welcome ideas comments X k X 35 THE STATUS OCLCt Thomas Hickey We have been running METAFONT on Tandem hardware since August 1980 This was accomplished by recoding the SAIL version into T TAL the systems programming language on Tandem The conversion took about six to eight weeks of con centrated effort plus another two to three weeks over nine months to add additional output modules and correct problems The system now is full version of METAFONT compatible with the original except in the following respects e TFM files are not yet generated e All of an identifier s name contributes to its uniqueness not just the first five characters plus length
56. b that try 1 will test whether 1 is empty or not If we give the input then the braces will be removed from this will be translated into empty_ tricka tricka trickb But in this situation the space indicated by is not ignored so ans will be defined to be F We might have arranged for the result of the bination empty tricka tricka trickb simply to be T or F rather than defining Vans to be T or F But if we did this construction like Mf T empty 1 tricka tricka trickb else wouldn t work because TEX would think that we were trying to compare T with the result of eapty 1 tricka The following variant of empty is also useful def emptygp 1 endd empty 1 tricka tricka trickb Then emptygp endd defines ans to and F otherwise is empty or D it will be convenient to use the flag ans for the result of several of our macros This won t produce problems if we ever have to perform two tests on two different arguments we can always first use enpty then let firstans ans then use esptygp etc 52 We also want to be able to check if is single token or group rather than string of several tokens or groups One idea is to consider single endd where single 1 2 endd checks whether 2 is empty def single 1 2 endd enpty 2 tricka tricka trickb This won t quite work since might be some t
57. boat Without volunteers there would be no TUG Steering Committee The future for and TUG looks bright provided we can ease our growing pains with your help Before I step down from my soap box I would like te thank all of you who have made the TEX Users Group and TUGbost possible through your involvement The staff of the American Mathematical Society deserve special thanks for their hard work and patience ks k PROPOSAL FOR MACHINE INDEPENDENT TAPE INTERCHANGE STANDARD Patrick Milligan BNR Inc At the Implementors Workshop in May a committee was formed to propose a tape format suitable for machine independent and operating sys tem independent interchange of TEX source files The members of this committee are Nelson Beebe University of Utah Patrick Milligan BNR Ine Robert Morris UMASS Boston Susan Plass Stanford CIT The motivation behind this proposal is to provide means of submitting machine readable source to TUGboat and someday to AMS journals as well as a means of distributing and exchanging Macros and manuscripts To some extent the prob lem of tape interchange formats has been addressed by TUGboat in its ASCII card image format 80 characters record 100 records block The primary problems with such a format stem from use of the full ASCII character set The fol lowing potential problems exist e Not all computer systems support
58. caret for the superscript delimiter Second macro packages typically include several font dec larations Incompatible assignment of font codes makes it difficult for users to select an assortment of macros from different packages If font codes as signed in a macro file do not correspond to the fonts preloaded by some versions of TEX strange results can be difficult to explain There is no total solution to this problem but it can be minimized Macro 43 packages should come with documentation describ ing the fonts and font codes used When sending files to another installation users should remember that preloaded fonts differ from site to site A helpful convention in assigning font codes is to reserve up percase letters for user declarations and to let stan dard macro packages use other characters Patrick Milligan s DefineFont macro described below can be used to automatically assign available font codes 4 k Macros Editor s note In an effort to hold down expenses some of the more extensive macro packages in future issues of TUGboat will be published on microfiche with a summary or introduction to each package included in this column Authors of macro packages who submit their work for publication here are requested to supply such an introduction along with the camera copy of the package Because fiche 13 not as easy to use paper an attempt will be made to arrange
59. codes available for font 1 elsef Apply First to UserFonts gt FontCode Get font code font FontCode 1 Load font xdef 2 curfont FontCode Define macro to use font Apply Rest to UserFonts gt UserFonts Remove code from list The Apply macro is used to apply a macro to its argument when the argument is a macro also The trick is to fool TEX into expanding the argument before the macro is applied If a better way exists to perform this feat please send your solution to TUGBoat Usage 5 1 lt function gt to lt argument gt gt lt result gt where lt function gt is the macro to apply lt argument gt is the macro containing the argument to lt function gt lt result gt is the macro used to save the result i def Apply 1 to 2 gt 3 let Func let Setup dummy function xdef Eval xdef 3 Func 2 Expand argument let Fune 41 Redefine function to use macro Eval Apply macro to its argument The First and Rest macros are used to manipulate strings terminated with an exclamation mark def First 1 2 1 fj Returns first character of string def Rest 1 2 2 Returns string with first character removed The macro UserFonts describes the set of font codes available to DefineFont The list of font codes should not contain an exclamation mark since this is used to terminate strings passed to the Fir
60. conserve memory space then overlaid on output multiple copies output of selected pages deletion by submitter of job from the queue and other similar features are also supported We have acquired the capability of generating fonts for the Alphatype via METAFONT see ar ticles by W J LeVeque page 39 and Ron Whitney page 40 The Alphatype is being used for limited internal production Another composition system is still being used for our major journals pending completion of necessary work on symbol and math fonts The entire two column portion of this issue of TUGboat and the separate address list have been generated on the Alphatype number of articles were sent in mag tape from VAX and Univae 1100 systems as well as from DEC 10s and 206 this experiment has been most successful and we hope to receive many articles for future issues on tape Some files have also been received via a phone file transfer mechanism not a network We have prepared camera copy on the Alphatype for several other organizations that do not have their own high quality output device for these jobs TEX input was provided to us on tape and the out put proved to be virtually identical to the original 29 Requests for Alphatype jobs will be considered on time available basis for price and other details call or write to Raymond Goucher American Mathematical Society P O Box 6248 Providence RI 02940 401 272 9500
61. count for the fact that one was deal ing with a binary rather than an ascii text file the same time the general file handling procedures TUGboat Volume 2 2 41 july 1 1901 00 09 10 1 Page 1 Cyrillic uppercase 3 Varian proofmode at the AMS J E Ad character description Cyrillic letter Eprime call charbegin 037 3codeoffset 11 0 0 phb O rbowl top3y2 3hh if bot3y2 lt ihh new y2 bot3y2 ibk 11 1ft3x2 round 75u 3 draw 2 lower bulb hpen lftOx3 round 1 25u y3 good6 8hh 1ftOx4 1ft312 4 2 20 8 rt4rx3 rtOx1 yi y3 topOy20 hh x5 5 x3 x7 u x6 5 x4 x7 topOy5zhh o 0 0 6 rtix7 round r u yT y8 Varian output of magnified character nus Imm 2 4 if ucs 0 0 ddraw 1 20 3 20 upper serif rpen w4 draw 3 0 1 5 1 0 erase spurious part fi hpen 0 draw 3 0 1 5 1 0 shoulder call a darc 5 7 w5 bowl y92y10 2 52hh x9 2u x10 1ft5z7 wO draw 9 10 bar 0 draw 6 1 0 4 0 1 tail 42 were changed to to allow greater flexibility in both specifying and reporting the full file names Changes were worked out first on an archaic TFX version of METAFONT but have since been transferred to a more recent TFM version in a rather straightforward manner Some modifications in approach have been suggested by work done at stanford although the d
62. countered our share of both in the short while the system has been up and running at Sandia If you fix a bug please send your fix to Barry The distribution tape is 9 track at 1600 bpi and consists of about 6500 blocks The tape was made using the command MCR BCK MTAO TEX BCKz data can be recovered by creating a directory TEX setting your default to that directory mounting the tape using MQU OVER ID MTAO and then recovering the files using MCR RST MTAO bck It is suggested that you put all the files on the tape in the TEX directory and initially use the program there The special Versatec driver LVDRIVER EXE must be copied from TEX to SYSEXE and installed This is done for Versatecs utilizing a DMA inter face by inserting two lines in SYSEXE STARTUP COM after RUN SYS SYSTEN SYSGEN but before AUTOCONFIGURE ALL The two lines to add are LOAD LVDRIVER DRIVER SYS SYSTEM LVDRIVER EXE and CONNECT LPAO ADAP 3 NUMVEC 2 VEC 0174 CSR 0777510 DRIVE LVDRIVER Reboot the system after these changes have been made The programs LVSPOOL and LHSPOOL spool output from th DVI file to the Versatec printer placing pages of output vertically or horizontally on the printer To function properly both programs need the privileges ALLSPOOL and PHY 10 When running from the TEX account the user must have these privileges if LVSPOOL and LHSPOOL sre run SYSEXE the system man
63. d ex bold and three width values condensed regular extended The Chelbegin routine then sets up the pens x height and other parameters which are used by the character routines None of the routines as sume that the vertical and horizontal raster resolu tion are equal The most important parameters are efactor and bfactor which control the expansion and boldness If a user is not satisfied with the range of values generated by the standard widths and boldness these parameters can be controlled directly characters are designed so that in nearly all cases they change smoothly with efactor and bfactor Exceptions to this include the tail on the lowercase a which has an abrupt transition when the bold ness is increased I estimate that designing charac ters to work over such a broad range takes several times the effort that designing a single font would Individual characters can be completed in as little as 15 minutes or may take several days average character takes 3 4 hours Rough routines for both upper and lower case have been completed but a great deal of work remains to be done to refine them and make them work correctly in all weights and widths Figure 1 is lower case b in proofmode Figure 2 shows this character in its 36 major variations 3 widths 6 weights and 2 slants Both of these were printed on an Anadex 9501 printer Completion date for this alphabet is somewhat uncertain although
64. d for the other architec tures Site coordinators are listed on the inside front cover and their addresses are given on page 2 If you have want current information please contact them Minutes respectfully submitted Robert A Morris Secretary Editor s note Attendees at the meeting may submit additions and corrections to the minutes in writing to the Secretary TUG TREASURER S REPORT June 30 1982 Beginning balance January 1 1981 419 Income Membership 1 555 Tape leasing 400 Workshop 7 445 9 400 Expenses TUGboat Vol 2 No 1 500 copies Printing 1 012 Postage 320 Clerical labor 60 1 392 Reprinting TUGboat Vol 1 1 300 copies 195 Vol 2 No 1 300 copies 655 Steering Committee luncheon San Francisco Jan 81 170 Workshop expenses 236 2 648 Estimate of future 1981 income Mernbership 100 members 1 000 TEX tape sales leasing 1 000 2 000 Estimate of future 1981 expenses TUGboat Vol 2 Nos 2 amp 3 800 copies Printing 3 200 Postage 900 Clerical labor 200 4 300 Reserve for 1981 expenses for Cincinnati meeting January 1982 1 000 Support for Stanford Coordinator 3 600 8 900 Subtotal 567 Anticipated receipts in 1981 against 1982 individual membership 50 of membership 4 500 Balance estimate to December 31 1981 3 933 Notes 1 Total membership is 495 of which 30 are com plimentary of these 371 me
65. ded for proof copy of text full math output doesn t seem practical on it at present Further tasks are to finish the conversion by finding out how to remove the temporary changes We believe these changes are due to incomplete im plementation of the system dependent routines We will also have to convert the remaining font infor mation files to Cyber format Our system dependent changes produce a NOS operating system version using 6 and 12 bit ASCII character codes for input and output we can find a way to test it we can build an 8 bits in 12 ASCII version for NOS BE sites TEX is a very large program around 98K words Our site only allows time sharing for 32K so we run as batch program While we realize that is intended to be interactive our CDC Cyber users will have to live with this problem until or we can find ways to make our version of TEX smaller If all goes well we plan to announce availability for CDC Cyber users at the VIM users group meeting in October We will need to write installation documentation and we would like to try out our installation procedure at a nearby CDC site If there is enough time before October we will do conversion of a more recent version of rather than the December 1980 version There are a number of attractive features in recent versions which would be helpful Several recently written Stanford programs are also valuable particularly one by
66. dered as casting indented just a bit more than the specified hanging indent Don explains it thus The word for is allowed after hangindent eg hangindent 3pt for 5 and moreover you can use letters as constants as in char e char 145 Then hangindent 3pt Forecasting means hangindent 3pt for 145 casting Likewise the word plus or minus will be gobbied after hskip ipt To disarm the trap braces to the offending macro 10pt 1 add a couple more Maangindent Probably most sites use version of TEX that has preloaded fonts and probably most sites preload the fonts specified in basic tex see the TEX manual Appendix B for details But some users for whatever reason decide to associate different fonts with the one character names assumed by basic When a file using non basic font names is run through a version of TEX that has preloaded basic the job may run to completion with no warning the SAIL implementation of TEX gives no warning but some Pascal versions may at least the VAX VMS version does so will use the metrics of the preloaded fonts but the spooler will use the fonts requested by the user and a ragged right margin may be only symptom TUGboat Volume 2 2 There are two solutions 1 back into the input files and change all conflicting font designators this can get very messy or 2 use a version of TEX in which
67. e If the first page in a DVI file had only a BOP and EOP command then the third page s BOP s previous page pointer TUGboat Volume 2 2 would be 46 since the first page s BOP command takes bytes zero through 44 the first page s EOP is byte 45 so the second page s BOP is in byte 46 When DVi reading program reads the commands for a page it should keep track of the current font This can be done with a single in teger variable the value of which will always lie in the range 0 2 2 The value of the current font is changed only by FONT and FONTNUM commands Whenever a command occurs in the file that causes a character to be set on the page the charscter is implicitly from the current font Likewise the program should keep track of the current position on the page The current posi tion on the page is like a cursor on the page whenever a character or rule is set it gets put at the current position on the page The current position on the page is just two numbers which are called horizontal coordinate and vertical coor dinate Moving to the right on a page is rep resented by an increase in horisontal coordinate while moving down is an increase in vertical coor dinate The upper left hand corner of the page is horisontal coordinate vertical coordinate 0 ie our system is slightly non cartesian Both coordinates are given in rsu s ridiculously small units where 1 10 meter
68. e digit text endwhile will produce text repeatedly until count digit becomes non positive Presumably text will set the counter non positive eventually whiles be nested if tLey use different counters repeat value times text endrepeat will produce text precisely value times where value can be either number or a counter The use of Rtemp repeat enables repeats to be nested to one level but no further For example repeat5 times x repeat3 times aA endrepeat endrepeat produces ii The fact that counter operations like multi pli ation and division are not provided by is one indication of their likely usefulness Of course that won t stop us from doing these operations anyhow def neg i setcount 1 count 1 det H1f 1 2 advcount9by 2 advcount9by 2 itpos9 advecount 1by 2 else adveount9by 2 advcount9by 2 def halve 1 savecount9 Htemp setcount9 count i advcount9 setcount 1 0 11 11073741824 17 1536870912 H1f 1268435456 H1f 1134217728 H1 167108864 H11 133554432 11 116777216 17818388608 17 14194304 17 12097152 H1 11048576 17 1524288 17 1262144 H1f 1131072 17 165536 11 132768 1 116384 11 18192 17 14096 H1f 12048 11 11024 1 1512 11 1256 1 1128 11 164 12 132 H1f 116 11 18 11 14 17 12 Hlf 11 setcount9 Htemp def m
69. e is a vertical box of height dimen containing a horizontal box of width dimen3 which contains hlist dimen be comes ivu Each position in the picture has coor dinates of the form z y where is the dis tance in vu from the left boundary and y is the distance in vu from the bottom of the picture Thus 0 0 is the reference point of the picture put something at position 36 475 simply type put something 36 475 The second coor dinate cannot be negative Both coordinates can be specified as the values of counters By putting puts inside puts a temporary change of origin can be affected allowing sections of the picture to be moved around in one piece For even greater flexibility picture a picture within picture The inside picture should be given width zero The overall scale of the picture can be adjusted by changing dimen Just for fun we ll give macros for inserting horizontal or vertical rules into amp picture and for drawing dotted lines def line 1 2 83 44 put setcount8 4 advcountSby 2 ifpos8 hskip 0 2pt wrule depthOpt widthO 4pt height count8vu else setcount8 3 advcountSby 1 wrule depthO 2pt width count6vu height 0 2pt 1 2 def speck hskip 0 3pt wrule height0 3pt depthO 3pt widthO 6pt def dotline i 2 3 4 5 put speck 2 3 setcount7 4 adveount 7by 2 setcounts 5 adveountSby 3 Divide i into7 Divide 1 intos setc
70. e were altering Pascal code rather than the macro language was written in The typesetter we are using is not addressable as a raster device and so we were unable to use standard fonts Instead we had to write a pre processor to convert descriptions of the typesetter fonts into format Various errors occurred in using s me of which were traced to errors in the TEX code others of which related to improperly formatted fonts The 1100 operating system allows a user to end terminal interaction with a program by typing an end of file signal TEX doesn t consider the existence of this concept Because we have no inexpensive proofing device we had to write a line printer proofing pro gram which has severe limitations due to the limited range of actions and characters possible on a line printer We have found the overfull box messages to be unhelpful in correcting justification problems and have replaced them with a more informative message We have found it very difficult to set type in narrow columns with TEX Details of all these problems are discussed below We should all hope that the project of transport ing TEX makes future language designers careful to be complete in defining language standards Hours and days of work could have been saved if the Pascal definition included external compilation the default case in the CASE statement etc External compilation in UW Pascal uses a syn tax very di
71. edgeable Acting in hastily convened and loosely organized telephone meetings we agreed to pay the cost of Mike s attendance at this meeting as our ob server even though TUG has no funds in its budget beyond those needed to pay for the newsletter Approximately 1000 will be borrowed from the AMS to be reimbursed from future TUG income In the Steering Committee meeting see my minutes page 4 it was agreed that we would propose no institutional dues until we had some proposal for use of such money to the benefit of the membership is such a benefit repre sentation of the TEX user community at standards committees and other organizations which may be in a position to influence the use or restriction of text processing systems for example I could en vision also presentations to governmental agencies who might be promulgating standards for govern ment documents Another benefit I think should accrue to paying institutional members is an annual tape of con tributed macros and perhaps a copy of AMS TEX when it is in its positive versions in the cur rent pre release versions I am enthusiastic about dis tributing it at cost to anyone who wants to test it Later I would make it a benefit of institutional membership Thus I now argue for the following dues structure Individuals approximately the cost of TUGboat Educational institutions 100 Non profit institutions using TEX in house 250 All others
72. egun to provide such a forum but in order to function effectively your as sistance is required At the time of the Workshop in May there were over 300 members of TUG It is not known how many of this number are actual users ss opposed to potential users awaiting work ing implementation of TEX on their local computer facilities In addition it is not known how many users have not yet become paying members of TUG By definition the TEX Users Group must have users of TEX in order to be a viable organization Therefore the primary goal of TUG should be to encourage and assist the growth of the user community There are several ways that you as a member of TUG can help 1 If you are lucky enough to have a working installation encourage your local users to join TUG In addition share your experiences with the use and or installation of TEX by sending letters articles bugs and macros to TUGboat 2 If you have received a version of and are in the process of installing it on your local com puter let TUGboat know about your progress or lack of progress News of temporary failure is just as important as news of success 3 If you are waiting for a version of TEX to be available on your flavor of computer architec ture contact your site coordinator to indicate your interest In this way you might be able to receive advance notice of a working TEX Also you might begin to acquire the nece
73. enneth University of Oregon TUGboat Volume 2 No 2 Rushworth Tom Block Brothers Industries Sachs Jonathan independent contractor Samuel Arthur Stanford University Schechtman Marty Newline Graphics Scott Eric P California Institute of Technology Sears Chris San Francisco State University Sherrod Phil Vanderbilt University Smith Barry Oregon Software Spivak Mike Stovall John Wycliffe Bible Translations Stromquist Ralph Univ of Wisconsin Madison Tal Avi Electis Engineering Incorporated Thedford Rilla Mathematical Reviews Trabb Pardo Luis Stanford University Truax Terry Mathematical Reviews Tuttle Joey I P Sharp Associates Van Dalen Gordon University of California Riverside Van den Bosch Peter Univ of British Columbia Wakabayashi Nobuo Stanford University Weening Joe Stanford University Welland Robert Northwestern University Wheeler Norman Whidden Samuel American Mathematical Society Whipple Edgar Lawrence Berkeley Lab Whitney Lynn Univ of California Santa Crus Whitney Ron American Mathematical Society Wilmott Sam Block Brothers Industries Wiser David Stanford Linear Accelerator Center Wolf Joe University of California Berkeley Zabala Ignacio Stanford University k kb xk amp g k PRELIMINARY ANNOUNCEMENT TUG MEETING CINCINNATI JANUARY 1982 The next TUG meeting will be held in Cincinnati Ohio at the Stou
74. ents table The equivalents table contains key information for all primitive control sequences and user defined control sequences This ar rangement is essentially identical to the SAIL version with the exception that the hash table pointers and macro text pointers are actually VM pointers The VM system used to hold the output routine alignment templates and mark texts is illustrated in figure 5 The addresses shown are not mandantory and can be tailored for a particular user s needs Additional VM systems can be added to expand allocated sizes of the various clements However the numbers shown are realistic This mechanism of fixed size allocation fosters simplicity and saves memory with little loss of generality Two output routine definition areas are shown allowing output routine redefinition without collision The next four byte blocks comprise the halign and valign template storage areas The remaining memory up to the 32K point is currently unused The second 32K is divided into 128 256 byte chunks Each chunk hold one mark Although there are only three possible marks per page of text top bottom and first passt does not know where the page break will be To solve this problem we keep up to 128 mark When pass2 decides it s time to digest the out put routine it indicates to pass how many marks were on the page just built Pass uses this information to manage the pointers and text buffer
75. ept identically defined but differently named types as identical in procedure parameters I was a bit confused by the INITPROCEDURE busi ness at first the documentation ought to say a bit more about this namely that that syntax allows compile time initialization on your compiler that it should be changed into a procedure in compilers without this feature and where it should be called in and TEXPRE We have a problem in the compiler with large case statements It does not handle statements with a large number of cases and the case statement in maincontrol gave some problems with this There isn t a fixed limit in the compiler but I broke the case statement in two and the compiler had no problem The sheer size of has given us some prob lems The UNIVAC s instruction set includes many instructions with a 16 bit address field that can only address 64K of data The data area for runs to something like 71K for us and we had to cut mem down from 32K to 25K to get the compiler to accept it This would have been easier if a Pascal version of UNDOC were available or if UNDOC had left memsize as a named Pascal constant instead of reducing it to 32767 and memsize 1 to 32766 etc had to go through with text editor and locate all references to 32767 and 32766 and deter mine by comparing the Pascal listings against the printed listings whether these were actually ref erences to memsize I seem to have
76. ercase CD Cyber Tab characters not allowed in the source P8000 Identifiers should be different in the first 8 characters VS UW P8000 Identifiers longer than 15 characters will not be accepted VAX No octal 20b notation VS P8000 All declared labels must be used VAX VS UW Can t take large procedures VS UW P8000 Can t take large arrays MULTICS No standard MAX and MIN functions P8000 Cannot take fields of packed records as actual parameters VAX VS Argument to PACK must be of type array it s not enough that it evaluates to array VS Loop counters must be local variables VS Labels and gotos must be local to same block cannot go to a label inside the else part of an if statement from inside the body of the true branch VS No nested WITH statements allowed UW Requires ENVIRONMENT modules for external linkage UW Variables must be initialized before their use They are not cleared by default VS No GOTO labels in enclosing procedures UW No INITPROCEDURE VAX UW MULTICS Can t take large CASE statements UW 17 No EXTERN procedures P8000 Instead of OTHERS the default case of CASE statements is ELSE P8000 OTHERWISE VAX CD Cyber OTHERWISE UW None MULTICS SUNY Can t pack memoryword properly CD Cyber In packed records elements defined of type 0 255 or 0 65535 are s
77. erform the reverse transformation and output a hex encoded tape The design decisions that went into the implementation of the system dependent module for Pascal TEX will be applicable to these tape utilities It is assumed that 9 track tapes will be used although the hex encoding would work equally well for 7 track tapes using a 6 bit ASCII subset for each digit The same coding scheme can be used to transfer files over phone lines if N is chosen to be a reasonable terminal line length An added benefit to this format is that it can be used to transfer binary data such as DVI TFM and font files with few modifications In this case the two hex digits would represent an 8 bit data byte instead of 7 bit ASCI character It seems clear that we need a tape standard that addresses the problems of machine indepen dent information exchange while still providing the functionality that requires There are two ques tions to be asked 1 Is this the format that we need 2 Is it worth the effort involved 12 Your input is needed to answer these questions Feedback from those of you who have been actively working on porting Pascal to new architectures is especially welcome Please respond Software THE FORMAT DVI FILES VERSION 1 David Fuchs TEX Project Stanford University April 18 1981 When TEX compiles a document it produces an output file that con
78. es of the implementating TEX in Pascal from having an in terest to having completed the installation so that there could be communication of the problems and solutions involved The first day consisted of a series of planned talks on various aspects of TEX from advanced usage to desirable features of Pascal compilers and techni cal details of TEX s output On the second day a series of informal sessions focused on people s prin cipal interests and concerns attempting to provide the information most necessary for those trying to install TEX and to gather the major unsolved prob lems hindering such installation Some of the more formal talks either appear as articles in this issue of TUGboat or will appear in subsequent issues Similarly number of the topics addressed during the second day have generated communications that appear here The range of in terests was large with the result that many par ticipants felt that much more could or should have been said about each of the topics Perhaps these communications will stir such discussion in these pages The schedule First day May 14th 9 00 10 00 Donald Knuth debugging aids detailed analysis of sample input using in formation available through features built into TEX such as trace and ddt It is hoped that a presen tation of this talk will be available for the next issue of TUGboat TUGboat Volume 2 No 2 10 00 10 45 Ignacio
79. evelopment has been inde pendent METAFONT description is device independent as well as point size independent within limits that limit is METAFONT itself with its highly device dependent routines and its inability to deal linearly with the full range of point sizes of interest to the typesetting community The METAFONT description may be device independent but METAFONT is not and that sort of generality is not possible without the use of translators to modify METAFONT s output The consequence of this last is crucial with respect to the creation of a font library and its implications are discussed else where in this issue The larger figure on the previous page is a Cyrillic 2 produced by proofmode Vertical unitlines and significant horizontal lines representing h height x height axis height baseline and descender depth from top to bottom form a grid on which the character is drawn Points mentioned in the drawing routine are labelled addition to proofmode which takes work to implement METAFONT contains a magnification option which is very easy to use n ordinary font production for the Varian one runs METAFONT and sets the variable pixels to 3 6 This value represents the number of pixels per point on that machine increased by 30 since we use the Varian for proofing To obtain the other character shown one simply fools the Varian into thinking that one point the printer s unit contains 3 6 15 pixe
80. f else construction chari and char are supposed to be single characters of type TUGboat Volume 2 2 0 to 12 or defined control sequences possibly with arguments that expand out to a character So we can t use an if else construction in situation where we don t know for sure what the next input text will be Suppose for example that cs 1 is supposed to be true text if 1 is a comma but false text otherwise If we define def cs i it i true text else false text there is always the possibility that our input text will contain Mes where is a token that can t be used with if or even worse a group which might produce total chaos In order to deal with this we will use several tricks which are also useful in other situa tions Basic Kludges Consider the definitions def false i gdef ans F def tricka A def trickb B def trickce 1 if iA gdef result false else gdef result gdef ans T result def empty 1 2 tricka trickc The control sequence trickce will be used only in situations where the if is safe In fact tricke will arise only from an occurrence of empty and the control sequences tricka trickb tricke and empty will be used only in the construction Venpty tricka tricka trickb Here will be some input text with perhaps few special 5 control sequences thrown but will never involve tricka
81. f the font divided by its design size or PST 15 just 1000 if there was no at specification for the font Next comes the byte fnamlen which is the number of characters in the font name followed by the the font name one ascii character per byte right justified Note that the font name includes amp directory only if the font is not in the standard default library directory From the definitions of the parameters of the PST command note that the end of the font definitions is marked by font number of 1 which is not legal font number The four bytes following this phony font number constitute the parameter q which is a pointer to the PST com mand i e the beginning of the postamble Next is a single byte parameter i called the ID byte Currently the ID byte should always have a value of 1 it will be changed to 2 on the next incompatible release of DVI format in 1990 Finally there is some number at least 4 of bytes whose value is 223 base ten 337 octal The idea of the q pointer at the end of the post amble is that DVI reading program can start at the end of the DVI file skipping backwards over the 223 s until it finds the ID byte Then it can back up 4 bytes read q and then do a random seek to that byte number within the DVI file Now the postamble can be read from start to finish while storing away the names and magnifications of all the fonts Now the program can jump to the start
82. ffer s Cincinnati Towers from January 11 12 1982 This meeting will review the growth and applications of TEX All TUG members are urged to attend There will be computer site dependent symposia as well as a general overview of We hope also to have demonstra tion of TEX preliminary schedule will be mailed to members early in the fall as soon as a program has been devised We would like to solicit reports on TEX implementation and usage Discussion topics which are submitted by September 15 will be con sidered for inclusion in the preliminary schedule Please send such requests to Tom Pierce Users Meeting P O Box 880 Collins Ferry Road Morgantown WV 26505 TUGbeat Volume 2 2 ASK WHAT TUG DO FOR YOU ASK WHAT YOU CAN DO FOR TUG Patrick Milligan BNR At the recent Implementors Workshop there were several discussions both formal and in formal concerning the future of and the TEX Users Group The following article reflects my opinions about where we should be headed and how can get there It seems clear that the widespread acceptance and use of is tied very closely to the success and growth of TUG Without an effective forum for the interchange of ideas and information will prob ably not fulfill its potential as a standard language for computer typography The TEX Users Group through TUGboat has b
83. fferent from that used at Stanford Any shared types variables procedures and functions must be declared in a separate environment module Only the head of procedures and functions is in cluded there the name parameters and result type comprise the head Each code module includes a list of procedures and functions which correspond to the heads in the environment To convert TEX into this format required a lot of hand coding Differences in I O handling required a lot of re coding in the system dependent module SYSDEP Of course it is to be expected that one would have to recode the system dependent code but it could be made clearer what part of the Stanford code refers to PDP 10 peculiarities and what is required by For instance TEX expects lines to be ended by carriage returns or line feeds etc which is the TUGboat Volume 2 2 internal file format of most PDP computers In the Univac file architecture carriage returns and so forth are not preserved in the file This must be the case with many computer systems and espe cially in Pascal compilers where such system specific things are supposed to be transparent What we did was to read a line from the Univac file and when Pascal returned true as the value of the end of line function EOLN SYSDEP appended a carriage return This would have been much easier had the SYSDEP code not also dealt with ignoring line feeds and nulls or rather had it documented and loc
84. gh the use of the chcode control sequence their placement and existence becomes critical Many powerful techniques presented at the mini course depend on the ability to redefine space or carriage return to invoke a control sequence Arthur Keller s nofill macro presented in TUGboat Vol 2 No 1 also uses this feature of TEX Another attribute of some text file repre sentations is limited line length worst case seems to be the fixed width card image format with sequence numbers Since al lows lines up to 150 characters unless care is taken TEX source may overflow the 72 charac ter limit imposed by some systems Even if conscious effort is made to limit line length there are times when it is difficult if not im possible to break a line for fear of introducing a significant space For example the qspace macro 5 has one line which is 98 characters long and it can t easily be broken since the space character has been redefined to be category 12 via chcode Many of the problems listed above must be resolved in the system dependent module of Pascal TEX for each architecture By definition our tape interchange format must be independent of the design decisions that were made for a specific im plementation of TEX The best we can do is provide a format that can be transformed into suitable input for Pascal TEX on a given system It is also hoped n that such a transformation
85. greatly facilitated had proofmode been avail able Other pending work involves the creation of new fonts and new symbols This work requires proofmode or at least some means for examining an enlarged representation of the character Proofmode was implemented originally for the Xerox XGP The routines for proofmode reside in the METAFONT module MFOUT Our task was to rework those routines to produce proofmode on the Varian which meant the translation of XGP commands to those suitable for the Varian or the Florida Data requiring the translation of absolute movement commands to relative ones That is for the XGP one could say move to column 545 whereas for the Varian one has to say move 224 units from last reference point The job was made more difficult by the fact that the XGP is not a available commercially and it was not immediately clear how to compare the XGP with the Varian The basic procedure was to make METAFONT proofmode write a DVI file containing all appropriate commands and typeset characters Full utilization of the DVI commands PUSH and POP to save and restore one s current position on the page prevented the need for the introduction of several real parameters to keep track of where one is on the page and the amount by which this position had changed This also prevented the occurrence of numerous accumulative rounding errors Changes were also necessary to the file opening procedures to ac
86. he National Institutes of Health in Bethesda Maryland At present it may used only by the DECsystem 10 staff Hopefully we will soon have facilities available that it may be opened up for general usage Currently there are four steps involved in running at First one runs TEX to translate text into DVI file Then the DVI file is used as input io program called DVIPDP DVIPDP essentially takes DVI records extracts all the important infor mation and outputs new records that contain op codes along with other information For example opcode 3 indicates a rule and the information that goes with amp rule are height and width Opcode 2 indicates a delta y item the vertical coordinate is moved by a certain number of pixels et cetera DVIPDP produces file of records like the ones described above This file is transferred to PDP 11 70 by means of a high speed com munications link Benson Varian printer plotter is hooked up to this PDP 11 A driver program TEX 11 was written for the 11 that takes PDP records and translates them into scan lines TEX at NIH has the ability to take graphs and figures and place them inside a TEX file TEX itself has not been modified to do this but the DVIPDP and 11 driver have the ability Graphs are created by using MLAB or OMNIGRAPH on the DECsystem 10 A program has been written that will convert a plot file into a file of alphanu
87. hing like token in this case 2 appearing in empty 2 tricka tricka trickb will still be empty since removes an outer set of braces from any argument So to be on the safe side we add some extraneous character after and let single 1 2 3 endd check if 3 is empty def single 1 2 3 endd empty 3 tricka tricka trickb Then single endd defines ans to be T if is a single token or group and F otherwise 69 Before using single endd it is essential to check that isn t empty Otherwise there will be problems because of the very considerations that made empty work An empty check could be incorporated into the definition of single but whenever AMS TEX uses single a separate check has to be made anyway D As in the case of empty a space may legitimately occur as argument 1 For example if we define def try i single 1 endd then try X defines Nang to be But try Lu defines ans to be T the second space never even gets read by TEX It is now fairly easy to check whether an argu ment 1 which might a priori be an arbitrary token or even group is a comma basic idea is to define def check 1 2 endd empty 1 tricka tricka trickb and then define def comma i check 1 endd that Ncomma 41 will define ans to be T if 1 is comma and F otherwise This won t quite work for the following reasons i If 1 or then comma i is
88. hole The Steering 58 Committee certainly wants its actions to be visible it wants to solicit the opinions of others and to encourage new volunteers However by the time all participants in such a large group have voiced their opinions it is impossible for decisions to be reached We need an effective decision making process The Steering Committee has also proposed rais ing individual membership fees and establishing institutional memberships This action has been delayed until TUG determines the services it will offer in exchange for such funds current situation illustrates both the need for some formal organisa tion and the need to raise money The ANSI X3J6 committee on text processing language standards is meeting June 22 26 Experienced users of two other mathematical typesetting systems have been invited to present the software with which they are familiar This ANSI committee has asked that TEX user also participate Although it is likely I will join X3J6 I am unable to attend the upcoming meeting Mike Spivak has volunteered to substitute for me but does not have institutional support for his travel ex penses The officers of the TUG Steering Committee strongly feel the Users Group should support this activity However our treasury is empty and it is not clear who can authorize such expenses The cure for this chaos is more work by the Steering Committee Sub committees should meet even electronically or by teleph
89. in Ohio 36 Local variable hiding Implicit in the design of METAFONT is the as sumption that users have the ability to add their own output modules for specific devices The METAFONT language offers no read access to the raster so such routines must be written in the implementation language to be included in your METAFONT system It is also possible of course to write programs which manipulate files in existing output modes such as CHR but this would not be a METAFONT program In my work in alphabet design I have found that much of the code is essentially defining two paths and filling in between them Successors to METAFONT will undoubtedly include the ability to define and manipulate paths as entities The con cept of pens and erasers is certainly useful but less fundamental than paths Conversion Problems One of the reasons TAL was chosen for imple menting METAFONT is its closeness in many ways to SAIL but there were a number of problems e ack of labelled case statements This was overcome by setting up arrays of ad dresses for jump tables e A limit of two to nesting of procedures Only a problem in the raster module Overcome by moving and renaming variables e Lack of SAIL s strings This necessitated the writing of a fairly com prehensive set of string handling routines e Lack of a macro facility as powerful as SAIL s Some sort of macro facility would seem to be needed for a clea
90. inate Actually the DVI reading program must have stack that can hold horisontal coordinates and verti eal coordinates as well as w x y and s amounts These six values always get pushed and popped together and a reasonable maximum stack depth might be about 200 times six since six items get pushed at once each page starts a DVI read ing program should set the amount variables to zero The stack should be empty The initial value of F doesn t matter since every page of a DVI file must have FONT or FONTNUM command before any command that will set a character the HORZCHAR and VERTCHAR commands Note that F is not pushed and popped program called DVITYP is available that takes any DVI file and prints a readable description of its contents together with error messages if the file is not in the correct format o Command Name Command Bytes Description VERTCHARO 0 Set character number 0 from the current font such that its reference point is at the current position on the page and then increment horizontal coordinate by the character s width 1 1 Set character number 1 etc VERTCHAR127 127 Set character number 127 etc NOP 128 No op do nothing ignore Note that NOPs come between commands they not come between a command and its parameters or between two parameters 14 PUSH POP 129 e9 4 p 4 Beg
91. ine count1 count4 count2 count4 elee ifpos6 line count3 count6 count3 count5 else advecount7 advcount1by 400 adveount 2by 400 advcount3by 400 endwhile advcount7 advcount4by 400 advcountSby 400 advcount 6by 400 endwhile endpicture B a a 8 EE NO n B lt E D TUGbost Volume 2 No 2 49 50 MACRO MADNESS Michael Spivak 2478 Woodridge Drive Decatur GA 30033 This article is an extract from the documentation for the not yet completed AMS TEX macro package It discusses certain tricks and pitfalls that other macro writers might want to know about Needless to say none of this trickery would have been possible without the help of Don Knuth should be mentioned that the 5 macro package initially chcodes the symbol 174 to be a letter and all internal AMS TEX macros contain a as one of their letters At the very end of the macro file is re chcoded to be of type 12 so that the 5 user cannot re define or even use these control sequences the in put will be read as For convenience we will omit the s here and we will use mnemonic names for control sequences the actual names used by 5 are very short at most three letters in cluding any Is in order to preserve memory space Please report any bugs to the above address as soon possible before the mac
92. inning of page The parameter p is amp pointer to the BOP command of the previous page in the DVI file where the first BOP in a DVI file has a p of 1 by convention The ten c s hold the values of ten counters at the time this page was output 130 The end of all commands for the page has been resched The number of PUSH commands on this page should equal the number of POPs 132 Push the current values of horisontal coordinate and vertical coordinate and the current w y and s amounts onto the stack but don t alter them so an after a PUSH will get to the same spot that it would have had it had been given just before the PUSH 133 Pop the s y and w amounte and vertical coordinate and horisontal coor dinate off the stack At no point in DVI file will there have been more than PUSHes HORZRULE 135 n 4 4 Typeset a rule of height h and width w with its bottom left corner at the current position on the page If either h lt 0 or v lt 0 no rule should be set VERTRULE 134 h 4 w 4 Same as HORZRULE but also incre ment horizontal coordinate by when done even if h lt 0 or w lt 0 HORZCHAR 136 c 1 Set character c just as if we d gotten the VERTCHARc command but don t change the current position on the page Note that must be in the range 0 127 FONT X4 w2 W3 W4 wo Y3 Y4 Yo 22 TUGboat Vol
93. is reversible An addi tional constraint placed on our tape format that it should be able to accommodate source con taining control characters significant trailing spaces and carriage returns and long lines It is not our place to pass judgment on the use of 8 some what esoteric tricks We must accept the reality that such features will be used In order to meet our constraints of machine in dependence and compatibility with idealized notions of text files we are proposing a tape format which represents source file as a stream of ASCH characters separated into lines by carriage return linefeed pairs This stream of characters will be broken into tape records N bytes long where N will be chosen such that 1 a tape record will exactly fill an integral number of words on all tar geted architectures and 2 N will be large enough to effectively utilize the tape Suggestions for a good value of N would be greatly appreciated The last block of the tape should be padded with NULs In order to avoid problems with helpful systems that like to throw away unwanted characters each ASCII character will be represented as two hexadecimal digits In order to make this format work each TEX in staller for a given architecture will have to write two programs One to read such a tape and trans form the data into a machine dependent text file format that will digest and another program to p
94. iversity Frisch Michael University of Minnesota Fuchs David Stanford University Gittelsohn Michael San Francisco State University Goldby Alan University of California Santa Crus Grosso Paul University of Michigan Guenther Dean Washington State University Hickey Thomas OCLC Incorporated Hodge Thea University of Minnesota Jackson Calvin California Institute of Technology Katagiri Grace University of California Berkeley Kelley Al University of California Santa Crus Knuth Donald Stanford University Lanford Oscar University of California Berkeley Lindsey Clark University of California Riverside Mapes Jeff Stanford University Melen Randy Stanford University Milligan Patrick BNR Incorporated Morris Bob University of Massachusetts Boston Nichols Monte Sandia Labs Norstad John Northwestern University Nussbaum Frank Newline Graphics Palais Richard Brandeis University Payne Thomas University of California Riverside Pierce Thomas EG amp G WASC Incorporated Plass Michael Stanford University Plass Susan Stanford University Platt Craig University of Manitoba Price Lynne BNR von Raesfeld Mary National Information Systems Reier Warren Gentry Incorporated Renz Peter W Freeman and Company des Rivi res Jim Carleton University Robb Richard Cemrel Incorporated Rosenschein Jeffrey S Stanford University Ross K
95. kes them harder to trace There s a new version due out soon as always e It s the old TpX in Pascal acquired from Stanford in late December Some of the bugs real TEX bugs too We ve just received the recent official release and as soon as I get a free weekend e There s amp bsolutely no documentation what soever that relates to the VMS version Gargon another weekend please The amaz ing thing is that this doesn t seem to matter Knuth s book describes the input format ex down to subtleties like tracing and the installation directions are rather concise put everything in account TEX TUGboat Volume 2 No 2 But when it s good it s very very good we re converting our skeptics to TpXnicians Details TEX in Pgscal for the VAX 11 750 and 11 780 running VMS with about 50 Computer Modern fonts in assorted sizes Comes with two spoolers horizontal and vertical for Versatec 1200 printer plotter using the standard Versatec interface Uses about 7000 blocks of disk space mostly for fonts and about megabyte whew of virtual memory while running Comes in source and binary executable forms sources for spoolers utilities etc ie we ll send you everything we have If you want to play with the Pascal pro grams you ll want the fancy listings available from Stanford copy of the Pascal manual mentioned above will be included on request Supplied only on magnetic
96. ls For the Stanford METAFONT this is done by specifying mode 1 to tell METAFONT that you want a Varian font and that magnification factor is coming and mag 15 The same sort of procedure would work for any other output device for which METAFONT can produce characters Although significant points are neither indicated nor labelled with this magnification option the figure obtained is of value in design Subtleties of strokes are more likely to stand out here than in our Varian output of proofmode It seems to us for ex ample that the vertex at the bottom of the letter is more apparent in the smaller than in the larger figure TUGboat Volume 2 No 2 F Warnings amp Limitations Xo Ro X amp Uppercase Update Fickle Fonts Last issue s warning about uppercase has now been rendered obsolete by a change in TEX whereby dimensions 5em and function words for after are recognized in any combination of upper and lower case letters 5EM For aFtEr This change was made in Don Knuth s own version of February 27 1981 and is described in the errata list among the Extensions since last printing But check the status of your local version before chang ing your macros Welcome though this change may be it lays another trap macro for formatting entries in an index included the specification hangindent iOpt 1 and an index entry Forecasting was ren
97. mbers are domestic and 124 foreign 2 The Implementors Workshop held at Stanford May 14 15 1981 was attended by 92 partici pants 3 Not included in these figures are costs for ser vices provided by AMS professional staff in cluding programming reviewing and editing answering telephone inquiries maintaining the mailing list and other clerical services 4 Professor Arthur Samuel is acting for Luis Trabb Pardo as coordinator answering questions distributing tapes and fixing bugs in the TEX source code Luis has asked and the finance committee has agreed that TUG con tribute to Professor Samuel s support Respectfully submitted Samuel Whidden Treasurer PROPOSAL FOR INSTITUTIONAL SUPPORT OF TUG Robert Morris Late last week June 14 the Finance Committee met with an unusual opportunity to fund some thing which has made me change my previous posi tion about TUG institutional membership Barring an obstruction due to an Air Traffic Controllers strike we are sending the chairman of TUG Mike Spivak to the ANSI standards committee on Text Processing Languages X3J6 This committee will be considering a number of possibilities for the processing of mathematical text and one of the Steering Committee members Lynne 6 Price will be attending as member However Lynne can not attend the beginning of the meeting and felt it important that be represented at least informally by someone knowl
98. me 2 2 In addition many typesetters come with their own font libraries which people would like to use A strong interest in these matters was shown by nearly everyone attending the TEX Implementors Workshop One result was the formation of a com mittee to investigate font related issues Members of the committee are George Ball Washington State University Barry Doherty American Mathematical Society David Fuchs Stanford University Tom Hickey OCLC Ron Whitney American Mathematical Society It has been suggested that this committee should first try to identify vendors who are willing to work with the TEX community to provide font infor mation for TEX s metric files and digitized rep resentations of fonts so that these fonts can be used with proof quality devices and to provide additional symbols for use with TEX The role cf vendors is complicated by licensing agreements royalties etc and inter vendor cooperation is un likely However the TEX community may be large and strong enough to have the potential for affecting this development positively and to its benefit It is probable however that such help from vendors will not produce fonts which can be reproduced on many other devices The need for a font library has also been pointed out either one maintained centrally or with some central means for pointing people in the right di ection Many tools are already available for manipulating font files e g transl
99. ment 2 is the first token or group following the comma and if it is a group the braces will be removed The removal of braces again causes problems if 1 is something like a frac b and also if 2 is something like a b where the braces are meant to hide the comma We could use braced here but it isn t quite foolproof since 2 might be a hidden comma which braced can t distinguish from an ordi nary comma Moreover braced can t help us with argument 1 Although this argument is usually a sequence terminated by a comma it just might happen to be a single group followed by a comma and there is no way of distinguishing between these possibilities once argument 1 has been read In the cases where 5 uses a recursive scheme of this sort the particular circumstances or simple tricks usually circumvent these problems The following definition illustrates a general scheme that will always work def kill i def op 1 2 endd ismarker 2 Mf T ans A kill1 1Z gdef nextop else A kill1 1Z gdef nextop op 2 endd nextop MdefNoperateiti 1 marker endd Notice that each time op 1 2 endd is used ar gument 1 now begins with which is removed by Vi 11 so it can t possibly be a group And argu ment 2 is always the remaining input terminated by marker so it can t be a group either recursive procedure can be used to count the number of commas in a string
100. ments is best illustrated by an example The macro options below will accept any number of single character arguments each of which will presumably cause some useful action If an x occurs it must be followed by two arguments which somehow belong to the x Also d implies a j as well The end of the argument list is indicated by a period A pos sible call would be options rx 30pt 75pt d Met options 1 def Next options Mf a 1 something else endif M j 1 something else endif MI def Noxt 1 2 something options else endif Mf d 1 something options j def Next options else endit Next A macro of this sort is invaluable in writing a general purpose macro package especially one to be used by many people A large number of different style options can be provided and each user can easily select any combination Research Problems 1 Speed up divide and sqroot 48 2 Write macro which tests two character strings for character in common Then dream up an application 4 Pictures In this section we describe a few macros which can facilitate the drawing of complicated diagrams The two macros at the heart of the method are these Mdef picture 1 2 3 4 endpicture varunit i vbox to 2 vss hbox to 3 4 hss def put 1 2 3 raise 3vu hbox to Opt Miskipi2vuitiNass A The result of picture dimen dimen dimen3 hlist endpictur
101. n venient There is also current interest in interfacing a general graphics capability with TEX As described in the last TUGboat Vanderbilt University has modified the Versatec spooler to allow output of plot files created in a format compatible with their Zeta pen plotter They intend to modify so that plot files can be merged with output Other installations are working on graphics extensions TEX s user interface particularly the input lan guage and error messages was also discussed as an area to be improved in the less immediate fu ture One specific point mentioned was the difficulty of identifying which spaces and carriage returns are significant Macro languages in general were criticized controversial suggestion was made that future systems be more like programming lan guages Joe Weening a Stanford student described his work on TEX derivative called La which is hybrid of and Lisp In LaTeX one can escape from into Lisp to do complex computations or text manipulations which are difficult or impossible to do in TEX Other topics included page markup and an inter active what you see is what you get version of TEX There was some discussion of feature that enabled users to tell where on a page material was being placed David Fuchs pointed out that such feature is incompatible with algorithm for determining page breaks TUGBOAT VOLUME 2 NUMBER 2 Con
102. n Road Evanston IL 60201 312 864 2896 WHIDDEN Samuel 8 American Mathematical Society Box 5248 Prowdence Ri 02940 401 272 9500 WHITNEY Ronald Amencan Mathematical Society Bow 6248 Prowdence Ri 02940 401 272 9500 ZABALA Ignacio Department of Computer Science Stanford University Stanford 94305 ZAPF Hermann Sestersweg 35 0 6100 Darmstadt Fed Rep Germany ZIPPEL Richard Massachusetts institute of Technology 545 Tech Square Cambridge 01239 TUGboat is the newsletter of the TEX Users Group TUG and is published irregularly for TUG by the American Mathematical Society P O Box 6248 Providence RI 02940 Annual dues for individual members of TUG are 10 00 for 1981 and 15 00 for 1982 one subscription to TUGboat is included Applications for membership in TUG should be addressed to the TEX Users Group c o American Mathematical Society P O Box 1571 Annex Station Providence RI 02901 applications must be accompanied by payment Manuscripts should be submitted to the TUGboat Editor Robert Welland Department of Mathematics Northwestern University 2033 Sheridan Road Evanston IL 60201 submissions to the macros and problems columns should be sent to Lynne Price CALMA Research amp Development 212 Gibraltar Drive Sunnyvale 94086 items submitted on magnetic tape should be addressed to Barbara Beeton or Barry Doherty American Mathematical Society P O Box 6248 Providence RI 0
103. n interesting 10 experiment in portability These examples just the beginning much more work needs to be done in this area Another direction for TUG growth is in the area of macro packages Most TEX installations quickly discover that one or more layers of macros are required to insulate their users from naked Many useful macro packages have been presented in TUGboat Michael Spivak s com prehensive 9 macros have been thoroughly documented in The Joy of However many more useful and interesting macros have been devel oped but not contributed to TUG Also the issue of portability is applicable to macro packages as well the use of extended ASCII character sets font codes counters and boxes all make the job of merging several macro packages together difficult Output device dependences may find their way into mac ros thus defeating device independent out put 16 is hoped that Lynne Price the macro coordinator may be able to bring some order out of chaos in this area with your help Awareness of the portability and modularity issues will assist TEX macro writers standards and conventions en couraged by TUG will also help Closely related to the issues of portable TEX support tools and macro packages is the area of machine readable distribution proposed standard for machine independent tape interchange is dis cussed elsewhere in this issue of TUGboat page 10 St
104. n translation and TAL s was adequate for most purposes e Restrictions on file names File names on the Tandem are limited to eight alphanumeric characters Output file names are constructed by adding a single character to the front of them so that CMR10 could produce files CCMR10 PCMR10 10 etc e 16 bit words This probably had the greatest effect of all The Tandem has an excellent repertoire of 32 bit arithmetic instructions but shifts and Boolean operations are only performed on 16 bit words For ease of implementation and run time efficiency a 3 word 48 bit cell was used for dynamic memory to correspond with SAIL s 36 bit word This does waste some space espe cially in the double cell section VMOM of TUGboat Volume 2 No 2 dynamic memory but is quite efficient since shifts and masks are not required to access fields In addition no new restrictions on sizes of fields in memory cells were needed e Different rounding of negative floating point numbers Rounding is now performed by a function call Chel am presently undertaking the coding of an al phabet called Chel This alphabet is based primarily on Helvetica and Helios and is designed to be us able over broad range of widths and boldness Chel is set up so that for each font to be generated amp call to Chelbegin is made specifying point size boldness and width Boldness can have any of six values extra light light regular medium bol
105. nal eof concept because the installation documen tation would say to have TerEof always return false at these sites Using TEX with a non raster device Our primary output device is Compugraphic 8600 typesetter The 8600 does use digitized data to create its characters but it will not accept user defined character shapes Therefore our TEX font files reflect the shapes of font characters available 32 from Compugraphic and our device driver outputs characters and commands to the 8600 rather than bit rasters We get an excellent quality of out put this way but the approach is not without its difficulties One problem was in converting character shapes into format The data had to be manually typed and a program was written to trans form the raw data into a pair of files one for and one for the 8600 driver Until all the bugs were worked out of this program mysterious TEX errors resulted from improperly formatted font files For example parentheses and other delimiters of vary ing size form a linked list in the font information At first our font preprocessor did not correctly indicate the end of the linked list This problem was difficult to diagnose because it caused a Pascal error TEX assumed the font information to be correct and so did not issue any warnings This is as it should be for maximum efficiency of TEX but it points up the usefulness of program such as the one desc
106. nce or interest in type design t 1 39 Also they commissioned Hermann Zapf of the world s outstanding type designers to design several fonts especially for the AMS which would be estheti cally consistent with Computer Modern Roman It was agreed that the augmented font family con taining not only Knuth s and Zapf s fonts but also a large collection of mathematical symbols already designed by an AMS staff member Mrs Phoebe Murdock for other purposes should be called the Euler family Dr Zapf has now prepared drawings of four alphabets upright Greek and Fraktur German upright script and an entirely new kind of alphabet which will be referred to as handwritten The lat ter is rather like an upright italic and will be used in place of Knuth s math mode italics Zapf s drawings have been distributed to the Font Committee and the resulting suggestions have led to modifications of several of the characters to accord better with what mathematicians are accustomed to The next step is to get METAFONT programs written to simulate these designs in 10 point type This work now being done by one of Knuth s graduate students Scott Kim should be completed within few weeks One of Knuth s METAFONT programs automa tically converts roman version of a character whatever that means for Greek or Fraktur to four others slanted sans serif bold and typewriter So these five variants will immediatel
107. no fonts have been preloaded such a version commonly known as VIRGIN TEX will start up much more slowly than a preloaded version owing to the greater number of font metric files that must be loaded at run time The following convention has been adopted at many installations Preloaded fonts use no capital letters Thus you are always safe if you introduce a new font called A B 2 Actually the AMS requires extended set of fonts including a full complement of cyrillic fonts 6 sizes these are called A F but G through Z remain open for special use Barbara Beeton L U M N Send Submissions to Lynne Price TUG Macro Coordinator Calma R amp D 212 Gibraltar Dr Sunnyvale CA 94086 The macro column is a new regular feature of TUGboat It is a forum where users can ex change formatting problems with or without solu tic 18 questions about writing macros comments on macros published in earlier issues of TUGboat etc Discussion of macros at the Implementors Workshop in May included some simple suggestions for increasing portability of macros across TEX sites First the excellent suggestion was made that ASCII Sites attempt to standardize the characters chosen to replace SAIL delimiters The conven tions are recommended ampersand for the tab character underscore _ for the subscript indicator and
108. nt four bits in every word are zero DVI file is actually series of commands command consists of one byte containing the command s unique number followed by number possibly zero of parameters to the command given command always has the same number of parameters These parameters may take from one to four bytes each but a given parameter of a given TUGboat Volume 2 No 2 command always takes the same number of bytes Some parameters may sometimes be negative in which case two s complement representation is used The complete list of commands with a description of all the DVI commands and their parameters is below The reader is encouraged to refer to the com mand list while reading the various examples in this document In the command descriptions lower case letter with a bracketed number following it means that the command has a parameter that is that number of bytes long An X3 command for instance is 3 bytes long the first byte of which has the decimal value 144 the second and third of which give the distance to move to the right If the second byte S and the third T then the distance to move is 28S T but if the high order bit of S is a one then the distance to move is 285 T 216 considering S and T as being in the range 0 255 The DVI file contains number of pages fol lowed by a postamble A page consists of BOP command followed by lots of other commands that
109. o be able to produce compilable code for all compilers automatically On the Univac 1100 amp user at a terminal can type eof to force an end of file condition in the terminal input to the current program The program must terminate without further intervention This possibility is not allowed for in TEX most likely be cause this concept of a terminal end of file does not exist on many computers We have circumvented this problem by having the InLnTer procedure in SYSDEP pretend that the user typed Vend if he indeed types an Geof This is not the ultimate solution since there are circumstances where will not accept a end command as for instance in the middle of a paragraph To help out with this problem and to make generally more friendly our basic macro file redefines the end command to be par vfill end This avoids the problem of an unfinished paragraph at the end but not of an unfinished halign etc The suggested solution to this is to have a function TerEof in SYSDEP to detect a terminal end of file condition On detect ing this a clean up procedure would terminate any current levels of processing possibly issuing warn ing messages Without having examined the code closely we don t know if the TerEof idea would be the best approach to modifying the existing code but the general idea would be the same no matter what the implementation This added code would not impact sites that do not have a termi
110. o the in stitutional memberships were the fear that such fees would inhibit the installation of TEX by small organizations or universities and b the current or ganization of TUG does not easily allow additional services beyond TUGboat as an enticement to make such fees worthwhile The bottom line seems to be that there aren t enough TEX installations willing or able to bear the burden of additional services such as TEX support or enhancement As the number of TEX in Pascal installations grows the direction and functions of TUG will grow also Once the first hurdle of providing TEX to a wide base of users is met there are other challenges for TUG to face In the area of output device support there is a strong need for portable device drivers and TEX support tools written in Pascal or some other widely used programming language Admittedly standard Pascal does not provide the full set of facilities required to write such device drivers but most Pascals provide some means of escape or ex tension to allow full use of the underlying operating system It would be a useful exercise in portability if large portions of device driver code were written in standard Pascal with architecture or operating sys tem dependences collected together in one or more system dependent modules like the SYSDEP code of TEX in Pascal One example of such a program is the Pascal version of DVITYP written by David Fuchs at Stanford Pascal TEX itself is a
111. of characters bozes are held together with a flezible space called gius After positional calculations are done the giue set average complexity text in about 5 to 10 seconds per page with hyphenation turned 08 Currently hyp henation degrades this by a factor of 2 or 3 but this is improvable The system mentioned has as its proces sor a Zilog 28002 16 bit microprocessor with 256KB of main memory and a secondary store consisting of a 10MB Winchester disk with an average access time of about 55ms The techniques described here are only an ex ample of those possible in the realm of software com pression The task of compressing software without gross performance degradation may not be a systematic one but this example illustrates its feasibility T REFERENCES 1 Knuth D E TEX and METAFONT New Direc tions in Typesetting Digital Press American Mathmatical Society 1979 CONTROL MAIN LINK m ud NN LIST SEMANTIC PROCESSORS STACKS 216 l SIMPLIFIED DATA FLOW SAIL VERSION CURRENT POINTER 10 EXPANS JON PREVIOUS POINTER FRAME HI WATER FIG 2 MACRO EXPANSION FRAME HASH TABLE EQUIVALENTS TABLE POINTERS TO MULTI CHARACTER SYMBOL NAMES PRIMITIVES AND MACROS manos aw SINGLE CHARACTER PRIMITIVES AND Fig 3 EQUIVALENTS TABLE TUGboat Volume 2 No 2 SYMBOL TABLE SYSTEM 64 HI WATER MACRO TEXT VM SYSTEM 64K BYTES GLOBAL
112. of the DVI file and read it sequentially The reason for reading the postamble first is that to figure where the characters on a page go the DVI reading pro gram must know the widths of the characters see the VERTCHAR commands description above To find the widths the DVI reader must know the names of the fonts so it can get their widths from or VNT or some other kind of font file But TEX can t put out all the font names until the end of the DVI file because new fonts can appear anywhere in the TEX job If font definitions were scattered throughout the DVI file then a spooler that read DVI files would have to read all the pages of the DVI file even if the user only wanted the last page printed The decision to put the font definitions in the postamble was based on these considerations and the fact that just about any reasonable sys tems language allows random access Unfortunately standard PASCAL does not offer this feature If it is absolutely necessary for a DVI reading program to be written in standard PASCAL then it either must make two passes over the DVI file or TEX must be doctored to output two files the regular DVI file plus PST file which contains only the postamble So far there have been no reports of any installation of TEX that required this kind of kludge 16 few words magnification If you document that does not mention any true dimensions then if yo
113. one between general meetings Someone must accept the responsibility of organiser and must be willing to make decisions even if they are temporary decisions later vetoed by vote of the entire membership I am as guilty as anyone else of neglecting my Steering Committee responsibilities except during meetings and the few days before TUGboat submission deadlines I for one will attempt to be more active in the coming months Sincerely yours Lynne Price x k k Dreamboat kd k kx amp Send Submissions to Lynne A Price Calma R amp D 212 Gibraltar Dr Sunnyvale CA 94086 One refreshing quality of the user com munity and particularly of the system s creator is that is viewed in fact intended to be the an TUGboat Volume 2 No 2 cestor of an evolving family of document formatters rather than as static piece of software that will be used for decades DREAMBOAT is a feature of TUGboat where users can describe in whatever detail capabilities they would like to see imple mented in some successor system brief Son of TEX session was held at Stanford in May Extensions of immediate interest include applications to non mathematical documents even those printed in languages other than English The foreign language application requires replacement of the English based hyphenation module For Hebrew and Arabic right to left formatting would be co
114. or programming this class of system on small processors It provided the best combination of both high and low level features of any language that had reasonable wide spread distribution The final major goal was to avoid using compres sion techniques such as interpretive systems Although interpretation widespread today most BASIC sys tems are interpretive and do provide for very com pact code it usually incurs a large performance penalty especially for programs requiring substantial computa tion TEX was expected to be computation intensive The original version of TEX was written in SAIL language developed at the Stanford Artificial Intelligence Laboratory for DEC PDP 10 s and 20 Furthermore TEX was a large program even with 36 bit words It seemed to have an insatiable appetite for memory while building pages of text The idea of actually compressing TEX into a 16 bit address space without using interpretive techniques initially appeared quite absurd 3 DEC and PDP 11 are trademarks of Digital Equipment Cor poration 4 Kernighan Brian W and Ritchie Dennis M The C Program ming Language Prentice Hall Englewood Cliffs New Jersey 1978 1981 TEBE Reprinted with permission from Digest of Papers CompCon Spring 81 February 25 26 1981 San Francisco CA The goal indeed seemed to a formidable task The methods employed to bring about the realization of Table top TEX is what this paper
115. ot enough to hold the numerous tables required by TEX most were moved to secondary storage and cached through memory resident buffers with buffer replacement done on a least recently used LRU basis We will refer to these as virtual memory VM systems How these VM systems are incorporated into various nooks and crannies of TEX will become clear shortly 4 PASS ONE The overall purpose of to break down in put text into a stream of primitives and data for pass who does the real work Passt is responsible for macro definition and expansion and for managing other user defined token lists such as the output routine align ment texts and mark texts These token lists are kept as character strings instead of hash indexes or primitive codes as in the SAIL version This provides for ease of passi manipulation whereas the SAIL implementation 381 TUGbost Volume 2 2 reduces them to pass 2 like entities Subsequently time honored string based algorithms can be used Figure 2 shows our macro expansion stack frame layout The hash entry symbol table is managed on a linear collision basis applying the hashing function to the symbol using this value to index into the hash table and then linearly scanning to find either a symbol match or a free cell Contents of the hash table are address pointers to the symbol character string see figures 3 and 4 The bash table is one to one with the first part of the equival
116. ount5 2 setcount 3 repeat i times advcountSby count7 advcount by count8 put speck count5 Ncount amp endrepeat Mine coords 4 2 will draw solid line between the points given These must be specified in the order left right for a horizontal rule and bottom top for a vertical rule TUGboat Volume 2 No 2 Mdotline value coords coords2 will draw a dotted line consisting of value 1 specks between the points specified which can be given in either order The last speck be misplaced by up to value vu due to rounding error so 1vu should be small if value is large dotline can be used to make solid diagonal lines by placing many small dots very close together but you won t get far before runs out of space Both line and dotline will only accept integer coordinates but this is no restriction if 1vu is small Mpicture can also be used as a very versatile and simple to use system for creating complicated symbols like d We conclude with couple of more complicated pictures Here is the source for the second def overt lower2 5pt hbox hskip 2 3pt u char 5 def cvert lower2 Spt hbox hskip 2 3pt u char 17 picture 0 083pt size size setcount4 5000 setcountS 4980 setcount6 4620 setcount7 0 while4 setcount3 3800 setcount2 3780 setcount 3420 while3 ifeven7 put overt count3 count4 else put cvert count3 count4 ifpos1 l
117. ply because they engender heated dis cussion The same attitude emerged in the schedule for the Implementors Workshop The program for the entire second day of the two day meeting was left unplanned in order to allow attendees to raise issues of their own interest With the limited amount of time available the breadth of the information to be covered and the number of opinions to be solicited it might have been better for someone to have taken the responsibility of making the decisions ahead of time The intent of the meeting was to provide demonstrations of output devices and discussions of implementations on various architectures These topics were postponed until the end of an intense conference While the other material was of unquestionable value it was of most interest to users who currently have access to and to in dividuals considering acquiring rather than to those who have decided to install TEX but have not yet succeeded in doing so It is ironic that Richard Palais pointed out that it has been over a year since general meeting of all TUG members surely had it been so advertised this meeting could have been one It is also ironic that Phil Sherrod suggested small workshops hosted by assorted TEX sites to describe their own installations Such a suggestion indicates that this meeting did not fulfill its intended purpose The May 14 Steering Committee meeting was open to the membership as a w
118. raphic languages e g Kanji symbols ff This permits large symbol sets to be used as well Zi evi gH The commands necessary to invoke font changes are simple character sequences which be generated by any host system without major software effort systems are however also capable of printing the output from Troff TEX Scribe and similar typesetting systems Fonts are software defined and can be local in firmware or downloaded from the host computer to the controller e Page Orientation Flexibility Portrait mode Landscape mode IMAGEN 12769 Dianne Drive Los Altos Hills CA 94022 415 949 1580 TUGboat Volume 2 No 2 IMAGEN CORPORATION page 2 of 2 Forms overlay Forms can be overlaid with whatever data is sent from the host computer to the printer A forms definition package allows for forms to be easily described in a high level language Logo and Signature Printing IMAGEN has the capability to scan and digitize hard copy originals in advance and store them in the host or locally in the system for printing upon user command In this manner logos signatures as well as other graphic images can be integrated into text output Sorter Collator With this feature reports may be transmitted from the host computer io the controller with the pages in any order The controller will then print as many copies of the report as specified keeping all pages in the proper order for each copy set without the need for fu
119. requirements of the growing membership Each member is entitled to receive all issues which appear during his membership year as well as Vol 1 No 1 If you have not received any issue to which you are entitled instructions for obtaining such issues are included on the form referred to above EDITOR S REMARKS Robert Welland We thank Lynne Price for taking on the respon sibility of editing our Macro column it is a complex task and we are thankful that it is in such talented hands In the future please submit all macros to Lynne A Price CALMA Research and Development 212 Gibraiter Drive Sunnyvale CA 94086 We also thank Barry Smith of Oregon Software for getting TEX up and running on the VAX see the VAX VMS site report page 34 and for making it easily available to all VAX users Because of this work we will see flourish at very many sites Due to the hard work of Thea Hodge and Michael Frisch of the University of Minnesota see their site report on page 28 we hope to see up and running on Cyber machines sometime this fall may the North Star guide them to success Lastly we extend the membership s gratitude to Barbara Beeton and Sam Whidden of the AMS whose hard work has made the TUGboat newslet ters possible x Editor s note Chairman Richard Palais is on leave for a year At the Steering Committee meeting m Moy Michael Spivak was ap pointed to se
120. ribed in TUGboat Vol 2 No 1 which automatically and correctly converts font files into readable data and vice versa Using a device like the 8600 limits the portabi lity of our TEX produced documents because the fonts do not correspond exactly to METAFONT fonts We have similar fonts for instance our English Times resembles Computer Modern Roman but slightly different widths can lead to different paragraphing and pagination overfull entries in tables etc We cannot use METAFONT with the 8600 because the manufacturer provides no method of addressing the typesetter as a graphies device and the cost of having fonts custom made by Compugraphic is prohibitive An interesting problem encountered in using a micro computer driven typesetter like the 8600 in conjuction with TEX is the lack of cooperation be tween the two devices The typesetter itself has fair degree of intelligence but it seems impossible to use any of it with TEX because of the need to make the DVI file device independent must assume it is dealing with an entirely stupid device The 8600 does justification tabulation automatic accent placement box drawing etc The commands we pass to the typesetter bypass all these features The question of cooperation between smart devices is a difficult one and it seems a shame that the smarts of the typesetter can t be better utilized This is general question not one addressed specifically at TEX tha
121. ro package gets dis tributed widely L Branching Mechanisms The only branching mechanism provided by TEX is Mf char char2 true text else false text and its relatives Unfortunately there are certain peculiarities of if else that require special care if else construction is processed in digestive system rather than in its mouth Suppose for example that we have two control sequences csa 1 and csb 1 2 taking one and two arguments respectively and a control se quence flag that is sometimes defined to be T and sometimes defined to be F We would like to define cs to be csa if flag is T and csb if flag is F the argument s for cs will simply be whatever comes next in the input text If we try to define def cs if T flag csa else csb then a use of cs will produce the error message Argument of csa has an extra because TEX sees the as soon as it looks for the argument after csa or csb The solution to this TUGboat Volume 2 No 2 problem is to define def cs if T flag gdef result csa else gdef result csb result similar problem arises the following situation Suppose that we have two different macro files mfile 1 and mfile 2 and the value of flag is sup posed to determine which file to use such a scheme is useful for saving memory space A definition like def cs if T flag input mfile 1 else input mfile 2
122. rrence of string may produce gdef newstring string but TEX will simply record this definition and not try to expand out the string that occurs in it until newstring is ex panded at which time an if test is made which produces a new gdef amp newstring should be defined as string rather than as string to keep internal input stack from growing unboundedly b Suppose that we have some input of the form stringi stringi strings with strings separated by some character like a comma and we want the control sequence Noperate to perform some operation on each string For ex ample we might want to replace each string by A string Z so that operate string strings stringn will produce A stringi ZA stringa Z A stringa Z We might also want to consider the case where there are no separators so that an A and a Z will be inserted before and after each token or group We will use the token marker as a marker to tell us when our recursion is over so we define def ismarker 1 compare narker 1 Now the basic idea is to define def op i 2 ismarker 2 Mf T ans A 12Z gdef nextop else A 1Z gdef nextop op 2 nextop def operate 1 op i marker omitting the commas in these definitions for the case of no separators Unfortunately this won t work because there are problems concerned with the removal of braces Each time op 1 2 is used argu
123. rtant because of the interaction between rules For example the horse prefix was put in not so much because we were concerned about hyphenat ing words like horse power correctly but rather to avoid hyphenating them incorrectly the rule 3 would break hor sepower The rules were mostly found by hand Good prefixes were found by looking through a diction ary suffixes by looking through a reverse diction Other ad hoc rules were discovered as the devel opment proceeded break vowel q break after ck However as good computer scientists we then used on line copy of the American Heritage Dictionary at Xerox PARC to test our rules This testing had two purposes 1 to determine which pairs of con sonants should be split under the rule and 2 to generate a list of exceptions to the rules The ex ception list originally contained thousands of words but was pruned down to just a few hundred Also in some cases new rules were formulated to take care of large classes of exceptions How well does the algorithm work in practice Quite well it seems Quantitatively in a test on pocket dictionary word list the algorithm found about 40 of the allowable hyphen points with about 196 in error Furthermore the hyphen points found were usually the most reasonable or good places to break the word In practice the algo rithm almost never makes a glaring mistake while 8t the same time the user does not
124. rther communication Business Graphics This feature allows the system to print curves of different shapes and widths as well as backgrounds intermixed with the text Business graphics features can be used to produce typical graphs charts engineering diagrams and in general drawings of medium complexity Full Graphies Capability This feature permits the system to generate an arbitrary pattern of dots so that any figure may be reproduced on the page In this way the system can be used as a facsimile receiver general plotting device and to do CRT screen dumping for computer graphics systems Text Setting Capability for Word Processors In typesetting applications the host computer usually generates the appropriate set of instructions to the controller to obtain the desired layout Text Setting capability allows for justification and other basic typesetting operations to be performed by the system This feature can be used to extend the capabilities of existing word processors most of which operate only in a typewriter style Communication Protocols The systems have the capability to communicate using network protocols This feature allows the easy integration of IMPRINT 10 as a node within a distributed environment File Management Capability An optional local disk and file management capability allow for font storage and printing job queueing This results in a self contained printing system in the sense that the host computer does
125. rtment of Mathematics Brandeis University Waltham 02154 On leave through June 1982 PHILLIPS Bob Oregon Software 2340 SW Canyon Road Portiand OR 97201 503 226 7760 PIERCE Energy Measurements Group Box 890 Colles Ferry Road Morgantown WV 26505 304 599 7585 PIZER Arnold Department of Mathematics University of Rochester Rochester NY 14627 716 275 4428 PLASS Susan 209 Center Information Technology Stanford Unwersity Stanford CA 94305 415 497 1322 PRICE Lynne A CALMA Research and Development 212 Gibraltar Orive Sunnyvale CA 94086 408 744 1950 SAMUEL Arthur Computer Science Dept Stanford Unversity Margaret Jacks 436A Stanford CA 94305 SCHWAB Rachel L National Institutes of Health Computer Conter Branch Building 128 Room 2 207 Bethesda MO 20205 SHERROD Pha Box 1577 Staton 8 Vanderbit University Nashville TN 37235 615 322 2951 SMITH Barry Oregon 2340 SW Canyon Road Portland OR 97201 503 226 7760 SPIVAK Michael 2478 Woodndge Dnve Decatur 30033 404 329 0372 TOVALL John 1500 West Comp Wisdom Road Dallas TX 75236 STROMQUIST Relph MACC University of Wisconsin 1210 W Dayton Street Madison WI 53706 606 262 8821 TRAGB PARDO Luis Department of Computer Science Stanford University Stanford CA 94305 WELLAND Robert Department of Mathematics Northwestern Urevarsit y 2033 Shenda
126. rve as temporary Chairman until Dick s return CHAIRMAN S REPORT Michael Spivak Since I am substituting for Dick Palais as Chairman of the TUG Steering Committee during the next year I suppose that J ought to emerge 4 briefly from the dimness of the 9 macro en gine room and report on the view from the bridge Up here it s all inchoate brightness everything s presently in a fog though there s the promise of smooth sailing ahead By the time of the Cincinnati meeting in January the official Pascal should be published and more important up and running at many more sites If you have encountered and solved any particular problems bringing TEX up your experiences will undoubtedly be of interest to others who want to implement TEX on the same or similar systems If possible please present your in stallation and or use experiences at a session of the Cincinnati meeting see the preliminary announce ment by Tom Pierce on page 8 Perhaps we ll soon be able to stop worrying about getting run ning and can concentrate on using TEX Two fun damentally opposed philosophies of how TEX should be supported were spelled out by Bob Morris and Sam Whidden in the last issue of TUGboat and it will certainly be interesting to find out just how much support is going to be needed since this will obviously influence the final decision Actually it seems that the problem of getting TEX running i e producing
127. s accordingly 5 PASS TWO The primary data structures of pass are large and complex linked lists For performance these lists memory resident However two tables are virtual in pas52 font information and the hy phenation exception dictionary Font information files typically range between 700 and 1000 bytes in length Since TEX supports 32 font files the need to these tables on secon dary storage is clear As before font information files are cached on an LRU basis with a small number of TUGboat Volume 2 No 2 files usually four in memory at one time Since font changes are relatively rare this causes little perfor mance degradation In fact the only VM system that has hindered performance significantly is the hyphenation exception dictionary Currently exceptions are kept in a file of sorted fixed length records and simply binary sesrched There are superior methods that we expect to incor porate later Perhaps the most important implementation decision in pass concerns execution speed rather than code volume The SAIL version utilizes floating point exclusively for its glue values This approach is un acceptable considering the poor floating point charac teristics of most 16 bit processors Moreover output results will differ between vari ous processors with slightly different floating point im plementations due to differences in accumulated round off The solution of course is to
128. s case the original definition of op should be made part of the definition of operate D There might appear to be possible confusion if some string contains within a group In AvS TEX this occurs only in constructions like align endalign where is temporarily re defined anyway Searching For Strings TEX s method of determining where an argument in a definition ends has the following peculiar fea ture Suppose we define def cs iab 2 Then the first argument is the smallest possibly empty token or group that is followed by a not the smallest group that is followed by ab So the input xayabc TUGboat Volume 2 No 2 gives the error message Use of cs does not match its definition So if we want to know whether ab occurs in some string we can t simply replace the comma by ab in the method of part IV d because an a might occur alone Instead we have to do something like the following def isb 1 compare b 1 def finda 1a 2 3 endd ismarker 2 Mf T ans gdef nextfinda else isb 2 Mf T ans gdef Hasab T gdef nextfinda else gdef nextfinda finda 2 3 endd nextfinda def hasab 1 gdetf Hasab F finda ia marker endd 4 g k xXx k k g Problems ks k amp The first formatting problems posed in this column come from the videotaped Class taught by Don Knuth last March Solutions will be
129. s in word on machines whose word size isn t evenly divisible by 8 BOP command has ten counter parameters size of rsu s has changed to be 10 7meter The postamble has changed to include overall magnification as well as a fraction that al lows use of non rsu dimensions Font checksum and magnification are new as is the convention about default directory name Font descriptions in the postamble give the length of font names rather than delimiting them with a quoting character The old zero ID byte is now a one Some ideas for version 2 Although 1990 is still a ways off we are currently expecting that version 2 of DVI files will differ in the following ways The ID byte will be 2 The q bytes of the pos tamble wil be preceded by s 2 where s is the maximum stack depth excess of pushes over pops needed to process this file SOME FEEDBACK FROM PTEX INSTALLATIONS Ignacio Zabala The Pascal version of TEX was designed and writ ten with the intent to generate a transportable pro gram Nevertheless given the characteristics of the TEX system some special assumptions had to be made about the Pascal environment in which PTEX was to be installed Essentially the requirements are The system should have enough addressable memory to store the large arrays employed by PTEX about 128K words of 32 bits The compiler should be able to really pack fields of PACKED RECORD and overlap multiple
130. sion of up and running and have no outstanding bugs Our interface to a 100pt inch Versatec is working satisfactorily and we are hoping to obtain the use of 200pt inch Versatec in the near future I am presently supporting at General Atomic at San Diego also our spooler only required a slight change to run there AND HYPHENATION Frank M Liang Word hyphenation is a useful feature of any com puterized document formatting system Sometimes it is also one of the most embarrassing The current hyphenation algorithm was de veloped by Prof Knuth and myself in the summer of 1977 Our goal was to come up with a reasonably compact algorithm that would find a significant per centage of possible hyphenation points but would make very few errors The algorithm is described in Appendix H of the manual Note that If you find any such embarrassing hyphenations done by you are encouraged to send them to the author 20 there have been quite a few minor changes since the original printing of the manual these are described in the errata file Basically the algorithm has four types of rules 1 Prefix removal e g com dis ex 2 Suffix removal e g able ful tion 3 Vowel consonant consonant vowel rule can usually split between the consonants and 4 Exception table about 300 entries Actually these parts are ap plied in the order 4 2 1 3 this order is rather impo
131. ss the next 1 to 9 tokens of input i or I to insert something x or X to quit because having a message like this implies that the host operating system will let the user type in both CR and LF and that it will distinguish between 19 them Some systems do n t do this either because they don t permit it at all or because it is not the normal way of doing things on that system Unix for example seems to turn both CR and LF into LF This problem cannot just be smoothed over in SYSDEP PAS because the help message above occurs in TEX PAS and because the procedure error in TEX PAS is the one which actually fondles the characters to see if we got CR LF PRINCETON PLASMA PHYSICS LAB PDP 10 HAMBURG PASCAL amp VERSATEC OUTPUT PHIL ANDREWS This is about the first thing I or anyone else here have done in Pascal and I had to guess at some of the differences between our compiler and yours It seems that TEX assumes that the loader will preset all variables to zero however our loader in serts junk some of the time Since our compiler doesn t have enough room to load in debug with it s particularly painful try ing to find errors Once I figured out how to bring up the first release I had little trouble with the others but 1 think some help could be given The major problem with compiling was the sheer size of TEX PAS and TEXPRE PAS which forced changes in our com piler As of May 9 I have the latest ver
132. ssary hardware for your output devices and begin to build some of the support software necessary to drive such devices 4 If no one is implementing on your flavor of computer architecture obtain copy of 9 Pascal and begin your own installation effort If you are not a systems programmer you should be able to interest someone on your local com puter staff to assist The intent of such communications to TUGboat is to minimize the reinventing of the wheel Each potential installer should be able to draw upon wealth of knowledge on the trials and tribulations of installation Each novice macro writer should have numerous examples available to learn from It is frustrating to hear second hand rumors at TUG meetings or workshops like So and So at SRI has working VAX UNIX TEX or Someone at DEC has a Diablo device interface or Somebody at has some nice thesis macros Just as Don Knuth has shared with the world it is imperative that you share your experiences with TUG Many of TUG s current problems are due to a lack of critical mass The porting of Pascal to many architectures and the availability of out put devices and their interfaces has not happened as quickly as anticipated At the TUG Steering Committee meeting in May the issues of institu tional memberships and TEX support were discussed but not resolved The primary obstacles t
133. st and Rest macros and it isn t valid font code anyway reasonable convention for font codes is to have 811 upper case letters available for user fonts def UserFonts ABCDEFGHI JKLMNOPQRSTUVWXYZ If DefineFont is used to allocate all fonts used including those in BASIC then all 64 possible font codes should declared 46 MACRO MENAGERIE Brendan D McKay 1 Math style testing One of the things which one should be able to do in but which apparently is impossible is to test for the current math style display text script or scriptscript For example how does one write a macro which produces a bold face g of the right def g hbox bf g obviously doesn t Here s a trick which doesn t completely solve the problem but which goes a long way towards that goal xdef subscr 4 xdef supscr t chcode 1 13 chcode 136 13 det us T dett 1 subser if MS T def uS S 1 else dez us x 1 deft 1 supser it us T def uS S 1 else der ms x 1 The idea is to maintain a macro MS which has the value S for scriptstyle X for scriptscript style and for all other styles including non math mode This can then be tested using the Mf macro The definitions above will maintain correctly if the style changes by use of the sub script or superscript characters but not otherwise Style change macros like scriptstyle can also be redefined to
134. t a few minutes discussing his plans for TEX which include a series of three books providing complete documentation on the system dates are projected completion dates entire listing of the Pascal source code a final user manual and a history of debugging TEX Winter 1982 Computer Modern Roman a description of this font family Spring 1982 METAFONT similar to the book Winter 1983 5 aaa a a Second day 15th 9 00 10 00 distribution and installation General problems of obtaining and of the transportability of both TEX and TEX related files Questions were raised about the real utility of the current means of distributing TEX in Pascal as two quasi independent documents Pascal source code and internal documentation both produced from T the same meta language source general opinion seemed to favor distribution of the original source together with the programs currently implemented only in SAIL for producing the pieces to allow each site to tailor the results to its and its compiler s needs more easily One result was the formation of a tape standards committee See the article by Milligan on this committee p 10 10 00 11 00 METAFONT and fonts Interest in both METAFONT and in the distribution of fonts Again one result was the formation of a com mittee to look into the problems See the article by Doherty 34 11 00 12 00
135. t in nicely with the review signed A Reviewer Ann Arbor Mich vskip 8pt noindent But sometimes an extra line must be added Vsigned N Bourbaki Paris To get This is case where the name and address fit in nicely with the review A Reviewer Ann Arbor Mich But sometimes an extra line must be added Bourbaki Paris 56 Problem no 4 Type 0 0 point 1 2 point 2 1 point 5 5 point 1 1 To get 5 5 1 2 2 1 e 0 0 e 1 1 TUGboat Volume 2 No 2 Problem no 5 Type hsize 20 End of a paragraph rightjustifythefollowing This is the first line it This is the second line 1 The third bf The last endrightjustify Beginning of another paragraph To get End of a paragraph This is the first line This is the second line The third The last Beginning of another paragraph Problem no 6 Type How do you do this lineskip 2pt baselineskip 1 3ex veenter halign hfil nfil cr linedown Look at this strange pile qquad veenter halign bfil cr lineup And at this stranger one To get How do you do this 8 o Poe wee ud TUGboat Volume 2 No 2 Balancing Columns of Text and Translation I would like to typeset translations in parallel with original texts using Perhaps there is a who can solve a forma
136. t of portability vs efficiency for a par ticular device One approach would be to have heirarchy of commands for instance a box com TUGboat Volume 2 No 2 mand in TEX which could also be expressed in terms of more primitive commands like HORZRULE and VERTRULE A user with an intelligent device could configure TEX to output the box command while other configurations could output 4 separate line commands To maintain transportability a stan dard skeleton for a device driver would have to be distributed which could break a box command down into 4 line commands at that stage The user with the intelligent machine would bypass this procedure Alternatively all configurations of a program like TEX could output the box command and distribute the rest of the work to be divided between the device driver program and the device as the programmer sees fit A Limitation of TEX We have found one apparent limitation to TRX it seems impossible to set text in narrow columns without extensive manual hyphenation One cus tomer wanted to set 9 point type in 2 inch columns When we ran we got many overfull box mes sages This leaves you to figure out the cause of the error It may well be that there is a word that TEX does not know how to hyphenate We have devel oped a non standard message for overfull horizontal boxes that prints the box as a single line on the ter minal representing characters as their correspond ing
137. tains specifications of how has decided the formatted text should ap pear in hard copy These output files are known as DVI files which stands for device indepen dent For instance running and telling it to input dviinf will cause TEX to look for a called DVIINF TEX read it and produce an output file called DVIINF DVI which is a DVI file This document describes the format of DVI files in detail giving all the specifications along with examples DVI file contains information about where characters go on pages The format is such that there are those who say that almost any reason able device can be driven by a program that takes DVI files as input In particular a DVI file can be printed on the Xerox Dover Xerox Graphics Printer XGP Varian Versatec Canon and Alphatype at the Stanford CS Dept depending on what spooler itis passed The DVI file is a stream of 8 bit bytes packed in computer words high order byte first If the com puter word length is not evenly divisible by 8 then the extra bits at the low order end of each word will be unused The first byte in a DVI file is byte num ber zero the next is number one etc For example on Stanford s 36 bit word machines byte number 0 is in the highest order eight bits of the first word DVI file while byte number 7 is in the twelfth through fifth least significant bits of the second word in the file and the least significa
138. tents July 1981 Addresses of Officers Authors and Others 2 Official Announcements 3 General Delivery Robert Welland Editor s Remarks 3 Michael Spivak Chairman s 3 Robert Morris Report on the TUG Steering Committee Sor d det tear E arr ordi 4 Samuel B Whidden TUG Treasurer s Report 5 Robert Morris Proposal for Institutional Support of TUG PNE Barry Doherty Report on the TEX Implementors Workshop Stanford 14 15 May 1981 6 Workshop Attendees Ux Sum S ESE 8 Tom Pierce Preliminary Announcement TUG Meeting Cincinnati January 1982 8 Patrick Milligan Ask Not What TUG Can Do For You Ask What You Can Do For 9 Patrick Milligan Proposal for Machine Independent Tape Interchange Standard 10 Software David Fuchs The Format of DVI Files Version 1 12 Ignacio Zabala Some Feedback from PTEX Installations 16 Frank M Liang and 19 Kent S Harris and Robert M McClure TEX on Small Machines 21 Output Devices David Fuchs Output Device News Flash APS 5 and Linotron 202
139. tions Found inconsistent definition and use of ReadFontInfo arguments Changed INITPROCEDURE to ordinary proce dure and deleted empty block at end of SYSDEP Changed OTHERS to OTHERWISE Changed type of brchar from INTEGER to AsciiCode This compiler does not allow GOTO labels in enclosing proced res in quit changed GOTO 100 by a comment Deleted unused labels Found nested WITH curinput in getnext twice MULTICS BENSON MARGULIES The compiler dislikes the construction INITPROCEDURE There is an array that is claimed to be too big May be solvable Impossible to deal with the need for an OTHERWISE statement which the compiler does not provide The filename interface of PTEX is still basically PDP10 oriented For a machine without a fixed number of channels the file opening interface is problematic requiring the establishment of an arbitrary limit U OF WISCONSIN UNIVAC 1100 82 U oF WISCONSIN PASCAL BILL KELLY A major problem in converting for the 1100 has been the differing methods of external compila TUGboat Volume 2 No 2 tion In UW Pascal all global declarations includ ing procedure and function heads must be included in an environment module It would be helpful if the same names were used for the same types in both and SYSDEP When we received TEX a type might be called packed hyphenbit in one and pckdhyphbits in the other Our compiler does not acc
140. together as if they were inde pendent probabilities which they are not and then compared to a threshold Adjusting the threshold obviously changes the performance of the algorithm One estimate is 40 hyphens found with 10 error In any case a large exception dictionary will be required for good per formance One reason for this is that looking at just four letters around the potential hyphen point is not sufficient The author has discovered examples where one must look as much as 7 letters ahead to determine hyphenation consider def i ni tion de fin i tive Patterns Currently the TEX project more precisely me is conducting research into better hyphenation algo rithms In particular I am investigating a method based on the idea of hyphenating and inhibiting pat terns For example a hyphenating pattern might be tion indicating that whenever tion occurs in a word we can hyphenate immediately before it Another good example is c c Note that hyphenat ing patterns are a generalization of the prefix suffix and rules discussed above In addition the idea of inhibiting rules has proved useful Such rules formalize the notion of we can usually hyphenate such and such a pattern except when it is followed by Also such rules are often useful for handling classes of exceptions More importantly we hope to be able to extract the rules automatically from an on line dictionary This will be done by
141. tored in whole 32 bit words Records are assigned to length of the longest freevariant possible P8000 All reports have received due attention Currently the code is all uppercase in Jines that are never longer than 72 characters All identifiers are shorter than 16 characters and differ in the first 8 charac ters Octal variables appear only in the module that contains the system dependencies Two more particularly interesting problems are worth mentioning here Eagle Berns while running PTEX with PASCAL VS detected a case statement for which no default had been provided and whose switch variable was out of range Intendedly execution should have resumed after the case statement and that is what Hamburg Pascal did PASCAL VS signalled an er ror Unfortunately this situation is left undefined in the Pascal report Bill Kelly using UW Pascal detected trouble in the statement pagemem curchar scanlength The function scanlength has the side effect of changing curchar UW Pascal opposed to Hamburg Pascal does not evaluate subscripts on the left side of the assignment until the right side has been evaluated The original SAIL program assumed that vari ables would be implicitly initialized to 0 and the as sumption was still valid for our Pascal Much work had to be put into initializing everything before its use Below we present a synthesis of some of the reports that have been most helpful in our project MOORE S
142. ts and spacing are needed and then produces these dots followed by 1 and the remaining input The removal of braces would be a minor annoyance if 41 were something like where the braces are meant to make the into mathord something that 5 users aren t supposed to know about anyway But it could be major catastrophe if 1 were something like a frac b To handle such problems we define def braced 1i empty 1 tricka tricka trickb Mf T ans gdef Braced 1 else single i endd Mf F ans gdef Braced 1 else gdef Braced 1 In other words braced puts back a pair of braces if 1 is or a group with more than one token or TUGboat Volume 2 No 2 group in it Thus Nbraced 41 defines Braced to be 1 except when 1 is token or in which case the outer set of braces is removed So aside from the unavoidable token case Braced has enough braces to give the same result as 1 IV Recursions There are several ways of handling recursions all of which are used at some point in AMS TEX Suppose that we want to define qns 1 so that Mns 1 is qms 2 is 3 etc We can define def qns i setcountl 1 def string ifposi adveounti by 1 gdef newstring string else gdef newstring newstring of def string string This only appears to violate the rule not to define a control sequence in terms of it self An occu
143. tting problem con cerning this type of typesetting It should be possible to recalculate the size of blocks taken by each language until both languages end on the same word at the bottom of their block Suppose that an initial estimate is made such that language A consumes 50 of the page and language B consumes 40 of the page Ten percent of the page is taken for margins When language A is at the bottom of its block language B has only con sumed 90 of its block By making the column of language A approximately 5 wider and the column of language 5 narrower the last word of both languages will more nearly come to the end of the block Is there an easy macro that will do this in Johnny Stovall k Input Dependent Macro Redefinition I would like a way of combining various non successive occurrences of certain types of input as the values of a macro For example initially we might define list 1 Then an occurrence of data in the input file should redefine list so that list 1 is while list 2 list 3 etc are empty Another occurrence of data sometime later should redefine list so that Mist dis Mist 2 is Mist 3 is empty ete etc Does anyone know how to do this Michael Spivak Dear TUG Members It was mentioned at our last meeting that TUGboat has yet to receive any letter to the editor submissions I would like to help rectify that lack by st
144. u change just its magnify statement the DVI file produced by TEX will change in just one place the word in the pos tamble that records the requested magnification The idea is that any spooler that reads the DVI file will multiply all dimensions the DVI file by the magnification thus the default magnification in the DVI file may be easily overridden at spooling time So if the document specifies magnifty 1200 a vskip 34cm will be recorded in the DVI file as 34 X 107 rsu s of white space but the spooler will multipy this by 1 2 making 40 8 centimeters of white space on output If the user tells the spooler to use a magnification of 1000 rather than the 1200 in the DVI file then the output will have 34cm of white space If a dimension in the document is specified as being true then divides the dis tance specified by the prevailing magnification so that when a spooler looks at the DVI file and mul tiplies by the magnification it gets back the original distance So if we vskip 24truecm while the magnification is 1200 TEX puts out DVI commands that specifies 20 centimeters of white space An out put spooler that reads this DVi file then puts 20 X 1 2 24cm of white space on its output Of course true dimensions will come out false if the spooler is told to override the magnification Font magnification goes one step further Assume for moment that the overall magnification is 1000
145. ultiply i into 2 setcount8 1 setcount9 count 2 setcount 2 0 while8 ifeven8 else advcount 2by count9 adveount9by count9 halves endwbile 47 def divide 1 into 2 setcount9 count 2 setcount 2 1 advcount9 whiled advcount 2by1 advcount9by 1 endwhile def Divide 1 into 2 ifpos 2 divideti 2 else neg 2 divide 1 into 2 neg 2 def sqroot i setcount9 count 1 adveount9 setcount i 1 setcounts1 while9 advcount9by count8 advcount 1by1 advcount8by2 endwhile halve digit will divide any counter other than counter 9 by two provided its original value is in the range 0 to 4294967294 Some of the earliest calls to H1f will need to be removed for machines with small word sizes sqroot digit will take the square root of any non negative counter other than counter 8 or 9 In the other cases the format is operation value into digit where value is number or a counter and digit is a counter num ber for the other argument and the answer value must be non negative in each case count digit may be negative for Divide or multiply but not for divide The restrictions on which counters can t be used and which counters are destroyed are most easily seen by examining the definitions Both halve and multiply are quite fast but divide Divide and sqroot take time proportional to the answer iii The method by which recursion can allow macro to apparently accept any number of argu
146. ume 2 2 137 4 Set current font to f Note that this com mand is not currently used by TEX it is only needed if f is greater than 63 because of the FONTNUM commands below Large font numbers are intended for use with oriental alphabets and for possibly large illustrations that are to appear in document the maximum legal number is 23 2 144 n 2 Move right m rsu s by adding m to horison tal coordinate and put m into x amount Note that m is in 2 s complement so this could actually be move to the left 143 n 3 Same as X2 but has 3 byte long n parameter 142 4 Same as X2 but has a 4 byte long a parameter 145 Move right x amount which can be nega tive etc 140 n 2 The same as the X2 command i e al ters horisontal coordinate but slter w amount rather than x amount so that doing a WO command have different results than doing an command 139 3 As above 138 m 4 As above 141 Move right w amount 148 n 2 Same idea but now it s down rather than right so vertical coordinate changes as does y amount 147 n 3 As above 146 n 4 above 149 Guess 152 2 Another downer Affects vertical coor dinate and s amount TUGboat Volume 2 2 73 151 3 74 150 4 70 153 Guess again 154 Set current font to 0 FONTNUM1 155 Set current font to 1 FONTNUM63 217 Set current font to 63
147. use fixed point with appropriate scaling There is a most conspicuous fly in this ointment It is inherent in TEX s line breaking and page breaking algorithms that glue values have the same dynamic range between 0 and 1 as between 1 and co This is essentially the definition of floating point The technique we chose is usually considered the worst possible solution software emulation of floating point The key however is the format used figure 6 At first a 16 bit exponent may seem a bit exces sive but since this is the natural width of arithmetic of most small machines it provides for rapid manipula tion Astonishingly enough has performed very well utilizing this technique The SAIL version uses floating point exclusively for all glue and dimensional data We limit the use of Boating point to gain both space and speed perfor mance With the exception of 32 bit glue values and line widths used within the line breaking code all of internal dimensional values are kept as 16 bit in tegers In this implementation of TEX internal units are mils 001 inch providing for a range of 32 767 inches For a standard 8 5 by 11 inch page this is certainly suflcient 6 CONCLUSION few words about performance are in order Actual measurements so far show that TEX an Onyx C8002 system can process 8 5 by 11 inch pages of 5 One of the more intoresting concepts TEX is the idea that characters and combinations
148. variants of packed records If this requisite is not satisfied PTEX will require at least four times as much memory The compiler should be able to handle large case statements say over 64 actual cases in a range 500 500 and have a default case this is non standard in Pascal but available in most compilers TUGboat Volume 2 No 2 Additionally PTEX requires an EXTERNAL separate compilation facility If no such thing is available the SYSDEP module has to be inserted both in TEX and in TEXPRE by hand Also if there is no compile time variable initialization the INITPROCEDURE appearing in the program has to be changed into an ordinary procedure Even though we tried to avoid it the fact that PTEX was developed and debugged on a PDP 10 with Hamburg Pascal influenced the way the pro gram was coded and documented This compiler was often permissive in the same way as other lan guages of common use in Stanford SAIL Only feedback from other installations can help us im prove the transportability of the program We have been lucky in receiving information from people who really worked on and reported both errors in the program and incompatibilities in the compilers The following are some of the problems that other compilers have had with the system As said it is often the case that the difficulty is due to the per missiveness of our Pascal and not to the installa tion s compiler Source must be all upp
149. veries in May 1981 IMPRINT 10 uses software definable fonts to print justified text in a variety of styles and sizes at high resolution Interfaces for the IMPRINT 10 exist which allow it to be easily mated to most host systems This page was printed on the system IMPRINT 10 is based on the Canon LBP 10 table top printer which evolved from office copier technology It prints on plain paper using a solid state laser and rotating mirror to synthesize images at a rate of about 10 pages per minute for 11 inch high paper with a resolution of 240 dots per inch PRINTER INTELLIGENT CONTROLLER SYSTEM IMAGEN C A general purpose display controller developed and introduced by IMAGEN Corporation with initial deliveries in May 1981 IMAGEN C accepts page layout information from a host computer and synthesizes the video needed to control the raster scanning printer or other display device It can be interfaced to one or more host computers using a variety of serial and parallel interfaces IMAGEN C contains diagnostic procedures for itself and accessible portions of its environment It is available only to OEM buyers with nationwide maintenance organizations FEATURES AND FUNCTIONS PROVIDED BY SYSTEM INTELLIGENCE Type Font Flexibility Suitably formatted files may be printed as justified text contain ing an assortment of type styles sizes and pitches including proportional The systems can handle extended character sets such as those encountered in ideog
150. very often need to specify explicit discretionary hyphens unless the columns are very narrow or letters very wide The algorithm takes about 4K 36 bit words of code including the exception dictionary note on the implementation If the algorithm is programmed by sequentially checking each of the rules to see if it applies it will run rather slowly Using hash table would improve things but a faster and more compact way is to use a version of finite state machine Interested readers should look at the actual code Time magazine algorithm This is reputedly the most widely used hyphena tion algorithm of acceptable quality The idea is to TUGboat Volume 2 No 2 decide whether or not to split word based on four letters wx yz around the potential hyphen point However this would require storing a table of 26 456 976 bits which is excessive Actually only about 10 15 of these 4 letter patterns actually oc cur in English words but it seems the storage would still be considerable R Instead the algorithm uses three tables of size 26 676 corresponding to the pairs wx xy and yz The origin of these tables seems to have been forgotten but they are supposed to represent the conditional probability of a break given that the first two middle two or last two letters respectively are a particular pair To decide whether to break at a given point the values for the three pairs are looked up multiplied
151. y be available for these characters and alphabets There is an additional step or rather there are many hundreds of additional steps required to yield what could be regarded as a complete collec tion of fonts in the Euler family As is well known to type designers type faces do not satisfactorily scale up or down in size in a linear way 10 point font reduced photographically to 6 points does not look right hair lines tend to vanish altogether thick lines get too thick upper case letters begin to dominate small lower case letters etc So it is necessary to ad just individually the various parameters that govern the relative sizes of these features for each desired size This is not a terribly time consuming job for a single alphabet but when there are five alphabets in five variant forms in ten or fifteen sizes plus all the mathematical symbols and their variations the job looms large This task is already under way in the Society office it is expected to be about a year before all the font tables one could reasonably ask for are available in the Euler family But long be fore then of course the ones most urgently needed will have been completed and it is hoped that TEX and Michael Spivak s macro package AMS TEX can 40 used for most of the composition work of the Society by the end of 1981 At first these fonts will be for the Society s own use but it is hoped that through licensing agree ments the
152. y publications it was apparent that additional fonts and symbols would be needed Here are examples of some of the gaps we saw Mathematical Reviews uses over 250 distinct non alphabetic mathematical symbols of which fewer than 100 are available from Knuth s font tables There is a large number of diacritical marks which are used in the various languages that Mathematical Reviews quotes references in and they occur on a larger number of letters than one might expect it is not feasible to reserve a place in a font table for each combination of letter with diacritic And of course mathe maticians put accents on letters in entirely un predictable ways If the accents are designed separately then the letters should be vertical if they are to center properly under or over the accents This requires for example an upright Greek alphabet Mathematical Reviews uses the whole Cyrillic family none of which had been designed Various publications such as the Combined Membership List are produced in very small type 5 or 6 point and these had been designed by Knuth only for sub and superscripts not for text What was needed was a text face as easy to read as that in telephone directory The AMS Trustees decided to create a Font Committee actually a subcommittee of its Committee on Composition Technology to work on this problem this committee contains both AMS staff members and working mathematicians who have had prior experie

THE T@C USERS GROUP NEWSLETTER EDITOR ROBERT

Contents

Download Pdf Manuals

Related Search

Related Contents