- Nottingham ePrints
Contents
1. Device Independent troff produces a low level device independent code for driving a variety of output devices We have christened this code DIC Ditroff Intermediate Code and its presence makes the interfacing of new output devices much easier than before A back end program has to be written to convert the DIC into the particular codes required by each target phototypesetter but with example code available from other driver programs this enables a working driver to be fabricated in about a day and a sophisticated one within a week The over riding requirement for any software running in the typesetter is that it must be at a simi lar low level of abstraction as DIC itself and should allow the setter to be driven in a slave mode That is to say it should support absolute and relative spacing commands at the full resolution of the machine Suitable primitives for point size change font change reverse area etc should also be supplied Any higher level software which for example insists on re justifying the text which troff has already justified or which limits the accuracy with which superscripts and subscripts can be placed should be politely but firmly refused Fortunately the T file language on the Omnitech and the Binary Byte language on the 202 have been at about the right level and have proved very suitable for our needs 6 Preview facilities One lingering problem that faces us is the provision of preview facilities at mo2. FOREWORD The attached paper appeared as part of a Workshop session of tutorial lectures held at the PRO TEXT I Conference in late 1984 The lecture was an informal presentation telling the story of how the University of Nottingham equipped itself just prior to the era of PostScript and desk top publishing to typeset all of its examination papers in house On a personal level my vol unteering to lead this project triggered an interest in computerised typesetting which later devel oped into a new direction in my computer science research interests This new direction has now lasted for more than 20 years and has grown to encompass almost all aspects of a subject now called document engineering The paper is of some historical interest if only to note that the equipment we used cost us the staggering sum of 80 000 in 1982 Nowadays comparable quality and much greater speed could be obtained from a computer and laser printer combination costing about one twentieth of that amount The key factors in the success of our project in those very early days were the willingness of the University not only to buy the necessary typesetters and computers but also to purchase a source code licence for the device independent version of troff This latter item at 4000 was not cheap but it has done sterling service and is still in use some 20 years later in the School of Computer Science Above all the device independent t
3. clusions By any standards our experiment has to be judged a success The savings made by our Uni versity in typesetting examination papers and other material in house will be sufficient to nullify the 80 000 cost of the PDP 11 44 computer and Linotype 202 phototypesetter after only two years We see further potential benefits in future from the UNIX DITROFF combination For example the UNIX environment would enable a database of student records to be kept on disc and at appropriate times items such as degree certificates registration cards etc could be pro duced on a range of output devices such as phototypesetters and laser printers So far there has been a lot of interest in the use of the UNIX phototypesetting software However the market for such technically orientated setting is rather limited and so there are few if any turn key systems available With this in mind it is clear that much of our success can be attributed to the fact that although we lacked some of the traditional composing and typesetting skills we did have to hand a fair body of expertise in mathematical typography mathematics and computer sci ence A system such as ours as it stands would perhaps be unacceptable to the trade where a more interactive wysiwyg approach might be favoured Nevertheless we feel sufficiently convinced of the merits of describing what you want in eqn notation rather than the more difficult task of maintaining standards and st
4. constant speed of light acceleration due to gravity Avogadro s number gas constant Rydberg constant Boltzmann s constant Bohr magneton Stefan s constant gravitational constant fine structure constant permittivity of free space permeability of free space radius of the earth molecular weight of air Yy c c for air o See wazna oa S lt gt 5 c w w 1 602 1918 x 10 C 0 9109547 x 10 kg 1 6726 x10 kg 1 6748 x 10 77 kg 6 626 1933 x 10 43 s 2 997 924 580x 108ms 9 8ms 6 022 098 x 107 kmol 8 3145 JK mor 13 6058 eV 1 380 50x10 JK eh 2m 5 6703 x 107 8W m K 9 27410x10 JT7 6 673 x 107 N m kg 2 e zj n 137 035 96 8 85419 x 1071 Fm 4r x 107 Hm 6 378 x 10 m 28 8 g mol 1 40 Figure 1 Front page of a sample examination paper This approach marks out a rather different philosophy than the one that is implicit in the original program Perhaps the only way to reconcile the two is for eqn to be re written so as to externalise as string definitions or as tables all the parameters which might potentially need to be adjusted to accommodate a given house style But it has to be admitted that attempts to make eqn more inde pendent of the ultimate output device are not as easy as would appear at first sight Ideally of course eqn simply feeds its output into troff which ultimately drives the phototypesetter In prac tice many of the de
5. ction 22 January 1979 Les1979a M E Lesk Tbl A Program to Format Tables UNIX Programmer s Manual 2 Section 10 January 1979 Ker1975a B W Kernighan and L L Cherry A System for Typesetting Mathematics Comm ACM 18 3 pp 151 157 1975 Cha1954a T W Chaundy P R Barrett and C Batey The Printing of Mathematics Oxford University Press Oxford 1954 Ker1982a B W Kernighan A Typesetter independent TROFF Comp Sci Tech Rep Bell Labora tories Murray Hill NJ March 1982
6. dest cost for our operators so that the final phototypeset effect can be judged and proofread before committing it to bromide We have had some limited success in this area using DEC VT102 terminals equipped with Digital Engineering Retrographics units but reliability problems and the slowness of the processor on the Retrographics board have held us up More sophisticated preview facili ties are available under UNIX System V on the BLIT now Teletype 5620 terminal but these are still rather expensive at the time of writing 7 Operator experiences The handful of people who control the keyboarding of the examination papers have all had a good educational background but do not possess any particular mathematical skills We were fortunate in the early days to be able to recruit a person with some experience of computers typ ing and the publishing trade But with only a few exceptions even the later inexperienced opera tors have learned very quickly how to transform an initial handwritten or typewritten version of an examination paper into a mixture of text troff commands and eqn commands The originals for the paper are marked up by our technical editor in troff and eqn notation but only to give occasional guidance where a rather abstruse combination of commands is needed to achieve the desired effect It can be noted at this point that a potent advantage of the troff eqn embedded commands approach is that standards can be imposed and maintain
7. e and leaks in the plumbing for the toner ink nearly drove us to distraction Fortunately this machine has now been withdrawn from the market to be replaced by the much more reliable Linotype 101 For our part we replaced the pilot system in November 1983 with a PDP 11 44 computer driving a Linotype 202 phototypesetter Although based on older technology the 202 positively gallops through the work and the 11 44 copes fairly cheerfully with driving up to six terminals and the 202 typesetter The Omnitech 2000 now put out to pasture in the Main Library is also driven from the 11 44 via a remote serial link 5 Interfacing 5 1 Hardware There have been few major problems here For reasons of price and performance we have purchased parallel interfaces on both our typesetters but we drive them down a conventional mul tiplexer serial line The necessary serial to parallel conversion buffering and parity checking is carried out by a Z80 single board computer The software for this interface was written using the Whitesmiths C cross compiler hosted on the PDP11 and LSI11 and producing code for the Z80 This code was in turn burned onto an EPROM mounted on the Z80 single board computer 5 2 Software After spending almost two years modifying troff you can imagine the software imple menters joy at discovering that a more modern and typesetter independent version of troff had existed all along This version Ker1982a now known as DITROFF
8. e organisations for typeset ting was in the region of 18 000 p a The situation was particularly aggravated by the require ments of the Mathematics degree examinations where rather than setting a small number of papers at the Final Examination an option scheme had been adopted which produced a separate two hour paper for each of about 50 possible options It will be no surprise to learn that the type setting and correction of these papers alone would occupy an experienced compositor for fully 6 weeks Somewhat nervously our University set aside 35 000 to purchase a pilot system consisting of DEC LSI 11 23 computer Linotype Omnitech 2000 phototypesetter and an AT amp T Phototypeset ter License An implementation team was set up with the following duties D F Brailsford project manager computer and phototypesetter interfacing initial version of mex macros D R Woodall installation and software management of troff tbl and eqn G F Paechter typographical specifications technical editor mex macro enhancements W J Armitage installation and maintenance of UNIX software assistance J P Onions with interfacing problems S J Marchant design of custom hardware with Z80 controllers to interface to parallel ports on Omnitech and Linotype 202 typesetters 3 Software development 3 1 Modifications to troff At this early stage we were still using the original troff program as supplied with Version 7 of the UNIX syst
9. ed If the paragraph indentation caused by a call of the PP macro is inappropriate or if an eqn command such as a sup 2 results in an improperly placed superscript then we know it is not a mistake on the operator s part but instead an indication of the need for an adjustment in the details of the PP and sup implementations 8 User reactions Later stages of proof reading are carried out by the technical editor and ultimately by the lecturer who set the paper There is widespread appreciation of the fast turnround of proofs espe cially since the commissioning of the 202 and there is broad agreement that out standards now meet or exceed those which we obtained when our papers were sent outside the University for setting There was some initial confusion among proof readers during the early stages of tuning the soft ware when an over zealous tweaking of some software parameter might cause a torrent of errors to appear in a paper that had previously been all right It proved quite hard to convince members of academic staff that the perceived errors were of a similar nature and that they would all disap pear simultaneously after a little more surgery on the mex macros or on eqn By contrast of course a more interactive scheme with what you see is what you get without macro assistance is closer to the traditional way of setting type but demands intense concentra tion to ensure that the house style is maintained 9 Con
10. em whose output was for GSI CAT machine A considerable amount of work went into rewriting troff to output codes which would satisfy the T file format required for driv ing the Omnitech 2000 A few extra features were also added in the form of new raw commands notably a mechanism for establishing modes such as bold italic or negative area Thus for example md B This is bold md B would change into bold face but would retain the underlying type style Times Helvetica etc This obviates the need to stipulate the explicit fonts being used and by re initialising the type style in use at the head of the document enables the same material to be produced in a variety of typefaces without altering the basic text A problem which this approach raises is what to do when a given style is not available for the font in use The solution adopted has been to produce simulated italic via the slope forward facility available on many modern setters Simulated emboldening is achieved by overstriking the character 3 times with a tiny horizontal displace ment on each occasion The effect this produces is just about acceptable for sans serif fonts but is of course no substitute for possessing the proper bold font The facility has been used in the main for emboldening certain mathematical characters that were not available in bold form on any of our fonts and which needed emboldening when used to denote vectors 3 2 The mex Macros An example of a
11. front page of one of our examination papers is shown in Figure 1 It will be immediately apparent that certain standard features such as headings underlines indentations of question numbers the phrase Turn Over on all pages except the last and so on can be set up as the standard house style A macro package called mex Macros for Exams was written and has evolved steadily over the past two years For example the first few lines of Figure 1 were obtained by typing EX 1144 EQ delim global cent EN UN FU S ce Physics sp ET 10 1983 ce PHYSICS HIII iii Sp STI 2 CM CR PH 3 3 Modifications to eqn Unlike troff the eqn preprocessor has only a rudimentary macro capability and the imposi tion of a house style becomes somewhat trickier Certain spacing defaults and actions are inherent in the implementation of eqn keywords but it is true to say that there is considerable scope for adjusting the spacing of elements within the equation by use of the and characters to denote full and half spaces respectively In principle this gives plenty of control over layout but means that each equation might possess many such marks to ensure a pleasant appearance Uncontrolled use of these spacing characters results in very diffuse and displeasing effects With the production of examination papers in mind our resident typographical experts decided to tune the eqn program with a view to imposing a com
12. g began when we received the UNIX operating system and began to run it on our DEC PDP 11 34 computer Quite apart from the merits of UNIX as an operating system we were intrigued by the text preparation packages nroff for low resolution terminals and printers and troff for phototypesetter output on a Graph ics Systems Inc CAT machine The nroff and troff programs Oss1979a have much code in common but we shall be concentrat ing on troff from now on The input format accepted by troff is the classic one of having layout commands embedded within the text These commands usually occur at the start of a line and begin with a character Other commands some of which are merely alternative forms of com mands can occur within a line and are preceded by a character So for example the text ps 10 vs 12 ft HR This should be in Helvetica whereas this switches back to f TRTimes asks for a point size of 10 and a vertical line spacing of 12 points followed by a change into the Helvetica Roman font Part way through the sentence troff recognises f TR as a request to change font to Times Roman The net effect is to produce This should be in Helvetica whereas this switches back to Times Troff is amply endowed with dozens of low level commands for asserting fine control over fac tors such as horizontal and vertical spacing hyphenation and the like To use the input language in this raw form requires considerable expertise and tenacity wi
13. ic and factual errors in the original paper The time taken to rebuild this paper over several lunchtimes was about 4 hours a large pro portion of that time was spent in ensuring the accuracy of the examination paper front sheet shown in Figure 1 In house Preparation of Examination Papers using troff tbl and eqn David F Brailsford Computer Science Group University of Nottingham NOTTINGHAM NG7 2RD U K ABSTRACT Starting in December 1982 the University of Nottingham decided to phototypeset almost all of its examination papers in house using the troff tbl and eqn pro grams running under UNIX This tutorial lecture highlights the features of the three programs with particular reference to their strengths and weaknesses in a production environment The following issues are particularly addressed Standards all three software packages require the embedding of com mands and the invocation of pre written macros rather than what you see is what you get This can help to enforce standards in the absence of tradi tional compositor skills Hardware and Software the requirements are analysed for an inexpen sive preview facility and a low level interface to the phototypesetter Mathematical and Technical papers the fine tuning of eqn to impose a standard house style Staff skills and training systems of this kind do not require the opera tors to have had previous experience of phototypesetting Of
14. ill translate these wishes into troff commands to effect them Tb and eqn follow precisely this philosophy The former Les1979a allows templates for tables to be set up and for tabulation commands to be given It recognises TS Table Start and TE Table End as being delimiters in the input text between which all typesetting instructions will be in tbl format Similarly eqn Ker1975a expects all commands between EQ and EN for displayed equations and between for in line equations in the eqn language As a simple example of what can be done the input 10 amma delta sub 2 over sqrt a sup 2 b sup 2 bh Re Wo fw a produces yt 52 la2 b2 with all calculations of the heights and widths of fraction bar and vinculum being performed automatically It goes without saying that preprocessors of this sort can outdo even the most demented human in terms of the sheer quantity fluency and dazzling obscurity of the raw troff output they generate 2 The Examinations Project Early in 1982 we were asked by the Examinations Unit at our University to assess whether a troff based system would be suitable for in house installation with the primary aim of typeset ting all examination papers over the crucial period from January to May every year At other times of the year a typesetting service would be provided for other University material At that time the annual cost of sending examination papers to outsid
15. much greater importance is willingness and flexibility in learning how to use computer systems 1 Introduction and Background Information It is a rather strange experience for an academic to be thrown into the role of storyteller and yet in this tutorial paper I shall attempt to convey the philosophy of the troff tbl and eqn soft ware running under UNIX and to relate our experiences in using it in a production environment at the University of Nottingham for the in house phototypesetting of a wide variety of material Although we have set posters newsletters journals pamphlets and forms using this software the majority of the work has been in the production of examination papers and I shall concentrate on this area in what follows It will soon be evident that setting up a system of this nature in less than a year and conditioning it for non expert use needs coordinated time critical teamwork of an order normally associated only with jugglers and acrobats From time to time those particular professions seemed strangely appropriate as we struggled with software bugs which insisted on replacing line end characters by strings of Ws or with phototypesetters which pumped toner ink all over our feet Nevertheless the results we have achieved make us feel it was all worth while and a tribute to the rest of the team appears in section 2 and in the Acknowledgements section In common with many other groups our interest in phototypesettin
16. pact equation style reminiscent of that found in Cha1954a Many default spacings associated with the centring of equations space around signs and so on were adjusted so as to minimise the number of explicit and marks needed when typing in the paper University of Nottingham FACULTY OF SCIENCE Physics PART IIB EXAMINATION 1983 PHYSICS HMI iii Time allowed TWO hours Candidates must NOT start writing their answers until told to do so Candidates may use a self contained silent electronic calculator in this examination provided that the make and type number are noted on the front of the script Programmable calculators must not be programmed prior to the examination and no program in any form may be taken into the examination room An indication is given of the approximate weighting of each part of a question by means of a bold figure enclosed by curly brackets e g 2 immediately following that part Marks will be given for the following where appropriate understanding the problem use of relevant equations obtaining an answer which is plausible making realistic estimates for those quantities that are not given for clear well labelled diagrams and for practical solutions to the experimental problems ALL questions may be attempted Total marks should be regarded as out of 80 The following data may or may not prove useful electronic charge electronic rest mass proton s rest mass neutron rest mass Planck s
17. roff unlike the original version of troff which was for the second generation GSI CAT typesetter only proved to be readily adaptable to a variety of typesetters and laser printers This meant in turn that the only other readily available software for device independent typesetting Don Knuth s TEX did not need to be considered Ruling out the use of TEX was no reflection on its considerable typesetting capabilities but simply a prag matic reaction to the fact that TEX needed a DEC VAX rather than a PDP11 and budget con straints absolutely ruled out this sort of expenditure COLOPHON As stated in the Acknowledgements section this paper was originally coded up in troff by Mar ion Windsor in early 1984 It was then typeset on bromide using the same Linotype 202 machine described in the paper The bromide was submitted to the Conference Proceedings as camera ready copy The source text of the paper was then lost during a transition from VAX based to SUN based UNIX systems in the late 1980s a sobering reminder of the importance of rigorous archiving policies This rebuilt form of the paper was obtained by scanning in from the Con ference Proceedings and then using Readiris OCR on the resulting TIFF files The paper was re typeset using UNIX troff to set up the correct typeface Times Equations and tables were re set using the eqn and tbl pre processors for troff and the opportunity has been taken to correct a few typograph
18. tails of the particular mathematical symbols available and their sizes will inevitably affect eqn s strategy Good example of this can be seen with integral signs J and square brackets It is clear from the source text of eqn that on the original GSI CAT typesetter at Bell Laboratories these are composite symbols made up from a kit of parts comprising verti cal and horizontal bars arc segments and so on On our Linotype fonts however we possess each of these symbols explicitly at 3 different sizes Heated arguments took place of almost theologi cal intensity as to where the overlap should occur in point size terms between magnifying a small version of f and diminishing a larger one and whether the stem weights were then accept able and so on 4 Hardware The pilot project from April 1982 November 1983 was conducted on a LSI 11 23 running UNIX and driving an Omnitech 2000 phototypesetter This system was successful in producing all the technical examination papers for 1983 Its performance was limited by the slowness of the LSI II even when driving only two terminals but even more severely by the slowness and unrelia bility of the Omnitech This particular typesetter made a brave and far sighted attempt to dispense with bromide and to use a laser driven raster scan system to set text onto electrically charged paper The overall technology is akin to photocopying in many respects but the problems of back lash on the drum driv
19. th the danger that the traditional typesetting technique of making fine adjustments by inserting lead spacers is replaced by the even more tedious drudgery of calculating such spacings in ems or in machine units Clearly forms of input are needed which express more abstractly and powerfully what the intended effect is to be This is accomplished in two ways Firstly troff has a macro facility which enables frequently used sequences of commands to be encapsulated in a single command name Files of standard macro definitions can be built up and two useful sets supplied with UNIX are ms for general typing of manuscripts and man for typesetting the UNIX manual pages Macro commands like the raw troff commands are preceded by a but the usual conven tion for macros is that upper case names are used As with macro systems in other areas of com puter software the power of a macro approach becomes especially evident when they can be sup plied with parameters or arguments which control the precise effect obtained Thus if the IP indented paragraph command of the ms macro package is invoked by typing IP hello 7 then the correct spacing motions will occur to set an indented paragraph with hanging tag hello and with an indentation of 7 ens The second method of making troff accessible to mere apprentices is to provide pre processors which allow the user to specify in a high level fashion what is required and which w
20. yle when repetitively doing what you want that we are happy to maintain the approach of tagged input to troff and eqn provided that we can have an output pre view capability of some sort before committing ourselves to the creation of film or bromide 10 Acknowledgements I should like to thank the Registrar and Deputy Registrar of the University of Nottingham for suppressing any misgivings they may have had and for supporting us in this project The Examinations Unit supervisor Betty Hickling and all her staff have remained stoically cheerful through many software changes floods of ink and the occasional bolt of lightning to the PDP 11 44 Brian Kernighan provided much helpful advice on his typesetter independent version of troff together with a fascinating technical report of work at Bell Laboratories in reverse engineer ing the Linotype 202 so that it could be driven as a slave device at a coding level even lower than Binary Byte Thanks also to Marion Windsor for typing this paper and for her hard work in the early days of commissioning the Examinations system I do not dare to add up the man hours contributed by those named in Section 2 but it goes without saying that the successful outcome of this project is a tribute to many late night sessions labouring over source code macro modifications and hard ware interface testing References Oss1979a J F Ossanna NROFF TROFF User s Manual UNIX Programmer s Manual 2 Se
Download Pdf Manuals
Related Search