Question & Answer Session with Donald Knuth
Contents
1. TUGboat Volume 22 2001 No 1 2 U K TUG Oxford Sunday 12 September 1999 Question amp Answer Session with Donald Knuth Philip Taylor I m sure everybody here by now knows Don Knuth Don has very kindly agreed to do a question and answer session this evening In order to record both the questions and the answers we have two microphones One for Don s use and a radio microphone so we can capture your question on the tape as well as the answer Many thanks Thanks Don Donald Knuth DEK O K I hope you can hear me Audience Use the microphone DEK No I can t use the microphone it doesn t work for me It s only for the tape but I will try to project Since I have been retired for a few years and my voice isn t what it used to be PI do my best A part of the rules are that when you ask me a question you give your name first Another rule is that I get to ask questions too Occasionally I did prepare one small thing because as com ing to Oxford I re read one of my favourite novels a Dorothy L Sayers mystery called Gaudy Night which is all about Oxford I wanted to read you just part of it which has to do with typesetting Laugh ter In Chapter 3 it mentions a Miss Lydgate who had prepared her magnum opus about I don t know how to pronounce it prosody the study of meters in poetry It said her handwriting was diffi cult to read her experience in dealings with printers was limi2. a constant in TEX I just left for later so that I could change it But you know originally I didn t think of having many users TEX was a system for me and my secretary so I wasn t thinking of things in much generality So the same mechanism that I had origi nally for category codes was also used for things like how many characters before breaking for a hyphen what was the penalty for widows et cetera all done with the same mechanism chpar C H P A R it was called Change Parameter I had no idea of gener ality in the first place Then later on as I saw the applications Math Reviews were making I realized that I d better have some way to allow more flexibility Jonathan you d be interested to know that the very first imple mentation of proto TEX was done without recogniz ing control sequences as words each character was read in as a character by itself and then the hash table lookup would go on afterwards The first im plementation which was done by my students in the summer of 77 while I was in China was very much like Active TEX in that sense We found out that we could make the program run a lot faster in its inner loop if we distinguished control sequences from ordi nary text but originally every character was active in the very first draft In this book Digital Typogra phy I resurrected the computer files that I had used when first getting my thoughts in order Dominik Wujastyk Don you made a chang
3. classes and it was taken over then by the New Yorker and other well edited journals until finally the copy edi tors who used Fowler s rules Fowler s book came out in the 20s I thinkt the copy editors who used Fowler s Audience Roberts rules DEK Whose rules Audience Just after Roberts rules DEK I never heard of Roberts rules of English only Roberts Rules of Order and I think your comment is out of order Anyway Fowler essentially gave an algorithm that pretty much boils down to this Look at what comes before the word which If it s a preposition or a comma then it s fine which can also be a pronoun But otherwise you change it to that Well it got to the point where almost all of the magazines in America and most of the newspa pers by the end of the 70s were using which versus that in Fowler s recommended style This was be cause the copy editors had risen through the ranks and won the battle I had written the first draft of The T Xbook in the old way but Guy Steele who was visiting Stanford from MIT marked it all up ev erywhere I had a wicked which Simultaneously my copy editor from Addison Wesley was revising vol ume 2 of The Art of Computer Programing which I had in TEX form before the 2nd edition came out I was doing the typesetting in 1979 and 80 and that copy editor was also a member of this new gener ation So that s when I learned
4. the algorithm for wicked whiches I could do a search with a text ed itor checking to see if it s following a preposition or a comma Soon I became very sensitive to this as were a lot of Americans I started to be irritated when peo ple would quote a sentence from my earlier papers where I used a wicked which It was even hard for me to read the New English Bible because I would silently have to translate which to that I go into 1 H W Fowler The King s English Oxford 1907 H W Fowler A Dictionary of Modern English Usage Oxford 1924 2 Somebody named Paul Roberts wrote Understanding Grammar in 1954 but I don t think it was terrifically influ ential 18 more detail in my book Mathematical Writing pub lished by the Math Association of America The British perspective I understand is different still but in America it s very much a stylistic lapse that calls attention to itself if you don t follow the convention You can go through now say 100 of the magazines in the last 10 or 15 years and you very rarely see a wicked which James Foster You said that the MMIXware book was for Springer did they ask you to use their ATEX 2 09 style files Laughter DEK I take it that ATEX 2 09 style files are a joke because that s an old version of ATEX James Foster We had to use them recently DEK I was given a free hand here But I m supply ing camera copy so we ll which
5. think by one of the noble men of his time He supposedly improvised a theme TUGboat Volume 22 2001 No 1 2 on that melody spontaneously but then he was fas cinated by it afterwards During the last year of his life he prepared a manuscript that he left unfin ished at his death called the Art of the Fugue That work is analogous to Perec s because the idea is to make a thing of beauty while working within tight constraints Sebastian Rahtz Good thank you Pause DEK Well if there are no more questions Laughter Elizabeth Gilliart What are you going to write next DEK Yes I tend to be writing about a page a day and so there s nobody alive that has read everything I ve written except perhaps me But I m just finish ing now a book that is incredibly specialised It will be in the Springer Lecture Series Lecture Notes on Computer Science and it s called MMIXware It s a set of computer programs to simulate the new RISC computer that I designed this year MMIX is a com puter for the new millennium The first M is for millennium and it replaces a computer called MIX that I used in my books on computer programming I had the privilege of working with the designer of the DEC Alpha chip Dick Sites who is one of my students John Hennessy the designer of the MIPS chip also was a participant in this design and a few other people in Silicon Valley We came up with something I think would be a fa
6. chapters instead It is the story of the people who live in an apartment block in Paris There are 10 floors and 10 apart ments on each floor and you go through the apart ments actually some of the apartments have sev eral rooms but you go through the apartments in the order of a knight s tour Eventually you find out about the lives of all these people and there are many other very interesting constraints that he put into the book Each chapter is a little short story kind of independent of the others Now you are asking if there s something else analogous in the domain of music or Sebastian Rahtz Film DEK or film O K The closest thing in film is this new movie Run Lola Run from Germany If you were to do it the way Perec did it you would have many more chapters but Run Lola Run gives you a story three times The first time ends in disaster and so Lola says No take me back Let s do it again So we start over and she does something slightly different in the first scene and then we go through the whole story again but everything happens five seconds later so certain accidents don t occur in the streets and the whole plot changes At the end of the second telling of the story it s another disaster not for her but for her boyfriend and that s too terrible to accept The third version of the story leads to a happier fate In music I suppose I think of a theme that had been proposed to Bach I
7. d these twelve fonts plus pothooks which she used to typeset So anyway you have some idea Does this mike actually magnify Can you hear me a little better O K Well hello I was reading from one of my favourite books but that s enough I have one more thing prepared then I will go for questions This was for Phil I wanted to show you Laughter amp Clapping Don shows he s wearing a T shirt under his shirt that celebrates the Aston UK TEX Archive oldest and best DEK I always try to wear the appropriate T shirt for the day O K your turn now Dave Pawson My name is Dave Pawson and I have an answer to your question I was brought up in the 1940s and 50s in the north of England and one of the houses that I moved into was built in the mid 18th century and around the coal burning stove and above it and beneath the mantelpiece there were a number of hooks and they were the pothooks for hanging the pots on The OED discusses this usage Does that answer your question DEK O K So this must be a similar shape that you would use for the symbols I think that someone said that shorthand uses pothooks some systems of shorthand Thank you Dave Pawson I think it s what you get when you type a left angle bracket into TEX and you are ex pecting a straight translation and one of your as sumptions in the big blue book comes unstuck be cause it comes out like a pothook Laughter Is that true DEK Migh
8. e important languages that Unicode supports Still the job of doing all of those in one system is so incredibly hard that I don t know where the ex pertise is going to come from to get it to such a well debugged level For me to put my imprimatur on something would require so much work I would have to check that the thing had been done right and there is so much involved in getting it done right As you know Unicode 3 0 which will be com ing out early next year really covers almost all the languages of everyone alive today they have filled in the last gaps they ve got Burmese and the Maldive Islands Sri Lanka the places where the political difficulties were they have several thousand extra Vietnamese and Chinese characters and so on And Yi and Mongolian and native American various Inuit and Algonquian languages are all there now But each of these languages has special difficul ties involved in the typesetting It s not just a mat ter of getting the symbols all kinds of ligatures and things must go in and rewriting of characters Many of the languages have no spaces between words and special hyphenation conventions and a total user community of a few thousand Moreover all the 19 people who do use some of these languages are ed ucated enough that writing in English gives them more job security thus the more they do to make it possible for everybody else in their group to use the system the less chance
9. e in TEX when you made it able to read an eight bit char acter and that was very important for European users and others I completely respect and like the idea that TEX is fixed that you ve finished with it as TUGboat Volume 22 2001 No 1 2 it were for the moment and that it is a fixed point and it s not going to change anymore However it does look as if TEX will have to go Unicode There is of course Omega already I ve never used Omega I haven t had the courage to take it on yet but I think it s there in my future and of course other people are using it I just wonder about whether it s going to just always be Omega or something like it and TEX and they re just going to become sepa rate things and go their own ways or could you give a Unicode version of TEX some imprimatur or some special blessing so that it becomes the one that ev eryone uses I wonder whether all the development efforts can somehow be brought together again DEK Yeah It seems so difficult I m a great fan of Unicode but I also know enough about it to know that it s incredibly complicated and that I would never have a system based on Unicode that I would be able to say has no bugs in it because of the extra complexity Well I like TeX I like having a pro gram that is if not 100 reliable it s 99 9999 it s as reliable as anything So solid that we can build on it This means however that I m not supporting all of th
10. hose things I suppose For some purposes of course these extra format restrictions are not only for your creativity but also for electronic archiving It s not the kind of restrictions that Perec would have added but I imagine the publisher sees that it s to their advan tage to have as many things in that style as possible It depends on what you are doing because that can also stifle what you want to say One of the things TUGboat Volume 22 2001 No 1 2 uppermost in my mind from the beginning of TEX was that I would have the freedom to introduce a new notation if it was the right thing for the sub ject I was writing about If I wanted to make up a new notation I wouldn t have to go through any middlemen who wouldn t understand the notation I would be able to typeset it and I would know that it was getting through in the way I wanted it Philip Taylor I have some bad news They really do want us out of here by half past ten so there s time for just a couple more questions Jonathan Fine I hope it s not too late to ask this question Why did you introduce category codes DEK Well if you look at this book Digital Ty pography it shows the original draft of TEX that I made the first night when I stayed up late typing a proposed design The feature was even more awful in those days it handled not only category codes but all kinds of penalty amounts and other param eters anything that I didn t know how to fix as
11. irly good machine to build in about ten years It tries to be the clean est computer design and easy to learn fairly nice to look at and to make theories about there is a group of people that are helping me use this new computer to rewrite the algorithms that I had writ ten for the old computer MMIXware is a set of soft ware programs that make MMIX live even though it hasn t been built yet The exciting thing to me is the pipeline simulator which is a meta simulator which means that it can simulate millions and mil lions of different kinds of possible implementations or even implementations that nobody knows how to build You can say for instance how much mem ory it has what kind of caches it has what kind of strategies the caches use to remember the recent in formation a pipelined computer has different func tional units you can say how many multipliers you have and you can define a functional unit to handle any subset of the 256 operation codes you can have any number of functional units you can issue any number of instructions simultaneously you can look ahead different ways and control all kinds of things And then you can find out how fast your programs run Now if we had that machine we could put TEX TUGboat Volume 22 2001 No 1 2 onto MMIX and see how fast it works with different kinds of caches This program for the meta simula tor is probably the most difficult computer program I ever wrote To me
12. is a special version of CWEAVE that makes little indexes on the right hand page One of the chapters in Digital Typogra phy talks about this Mini Indexes for Literate Pro grams is the title of that chapter Basically when you are reading a computer program every name of a variable on the page that you are reading is either defined on that page or you can find it in the mini index And the mini index will tell you that the variable is say defined in section 5 and it is a pro cedure name or a constant and so on It gives you a quick reference It s an idea that I picked up from some textbooks on languages where people would be learning French or Russian Given a short story in some other language every word in the vocabu lary that you might have to look up in a dictionary appears in an index on that page With CTWILL I have a bunch of macros that check the uses and definitions of everything on a page The macros au tomatically prepare mini indexes there is a separate pass to sort the words but the first pass does the layout Some of the more difficult probably the most difficult T X macros that I ever wrote are involved in that I call this program TWILL because it s sort of a double WEAVE Anyway the CTWILL program gives me the output that I have to send to Springer So no I won t be using their style files I did one other book in their series but that was some years ago before they had learnt to insist on t
13. it was most interesting because it de pends on the idea of literate programming that I used when I worked on TeX for me perhaps the greatest spin off of TEX was this idea of literate pro gramming which has helped me write computer pro grams of all kinds I don t think I could ever have written the MMIX meta simulator without literate programming it would have been too mind bog gling I would not have been able to get the whole thing together and debug it it would have been too much of a mental strain overburdening my head In the past literate programming has often helped me write better programs but here for the first time it was crucial or I couldn t have written the pro gram at all I don t think MMIXware would have been possible without a good documentation lan guage to help me understand what I was doing as I went along So this will be a book about 400 500 pages but it s mostly just the typeset version of these literate programs for that new computer Then also Ill be finishing a sort of sequel to Digital Typography Digital Typography is volume three of a set of collected papers all the scientific papers that I ve written are being divided into eight categories The first book contained the things I wrote about Literate Programming The second book was called Selected Papers on Computer Sci ence those were the papers I wrote for audiences that weren t primarily computer scientists it col lects the general w
14. orks And then the third volume was Digital Typography and I told Phil d come here tonight because TIl do anything to sell copies of that book The fourth volume will be called Selected Papers on Analysis of Algorithms and those are my mathematical papers for what I think is my main unique life s work the study of computer meth ods in a quantitative way How good are particular computer methods from a definite exact mathemat ical point of view That book is estimated to be about 750 pages and it will contain I think 37 38 papers on that subject The material has all been scanned and put into TEX form But I have to spend a few months in spare moments going through dot ting the i s and crossing the t s and making the in dex putting the bibliographies in a consistent for mat finding out people s middle names for the in dex things like that And I also go through every paper and put it into the form in which I would like it to be remembered So if a paper was written in the 70s and I used sexist pronouns I change that 17 I try to rework it so that instead of saying he did it PI say they did it or something Also I change which to that a lot It s an American thing Male Voice Why DEK It s actually because of an Englishman named Fowler who wrote The King s English and gave rules for which versus that The people in America be lieved him and they started teaching English
15. t be The other thing I wanted to say is thank you very much to Phil Taylor for arranging that we could have the 10th anniversary celebration during the brief window of time that I could be in the U K I don t get to travel very often and so now I get to remember not only the 10th anniver sary celebration of TEX Users Group in America but also the one from here At the 10th celebration in America we had the president of TUG dressed as 16 META the Lion that was Bart Childs and now here we have Mrs TRX as well so now my delight is complete Sebastian Rahtz On your recommendation last year I bought a copy of Life A User s Manual by Georges Perec which I am still trying to read I wonder while I m still trying to read it whether you would like to recommend a film or a piece of music which has equal meaning to you as Georges Perec s book DEK I guess when I made my home pages a few years ago one of the pages listed books that I was recommending to read There s this incredibly dif ferent book by Georges Perec called Life A User s Manual which is a combination of many different kinds of artistry It has a mathematical basis but still becomes I think a great work of fiction He de veloped this book with 99 chapters There should really be 100 chapters but each chapter was based on a lot of mathematical constraints and one of his rules was that you had to break one of the rules so naturally it has only 99
16. ted but she had invented a novel and com plicated system of notation that involved the use of twelve different varieties of type and then she had all kinds of sheets in page proofs and so on And she said Don t prick your fingers on that bit of manu script that s pinned on I m afraid it s rather full of marginal balloons and interlineations but I sud denly realised I could work out a big improvement in my notation so I ve had to alter it all the way through Then Harriet said comfortingly Well the Oxford University Press is no doubt accustomed to deciphering the manuscripts of scholars Laugh ter Now this work of Miss Lydgate appears to play a kind of minor role in the entire book and I have Thanks are due to the following Philip Taylor for tape recording the Q amp A session Jonathan Fine for managing the transcription Pam George transcribing from tape and making corrections Don Knuth for diligently and promptly correcting proofs 15 a question for you here because it said her system of scansion required five alphabets with a series of pothooks for its expression and this is a term I don t know What are pothooks If any of you can tell me Audience Spell it out DEK It s spelt like pot hooks P O T H O O K S Audience Pothooks are hooks for hanging pots Male Voice Yeah but in typography it s an S like an S DEK Is it some kind of special symbol O K So she ha
17. they have of being uniquely able to do anything So it s going to be hard to support this commercially it s surely going to be a volunteer effort The effort is not only an order of magnitude more difficult than what I had to do but it also has to be done pretty much as a labour of love as I did it So it looks to be a while before it could converge like that not that it s impossible but I myself wouldn t be in a position to bless it All I can do is provide an example of one of the world s nearly bug free programs so that other people can try to emulate the good points and correct the bad points Well thank you very much Applause Philip Taylor Don I d like to thank you very much indeed on behalf not only of the Commit tee of the U K TEX Users Group but of every one of the members here who I m sure are absolutely delighted that you have spoken to them Thank you very much indeed for your time for joining us Jill thank you very much indeed as well for coming along It s been a great pleasure to have you both in our midst We wish you a very safe and happy stay in the U K for the rest of your trip Thank you very much indeed _ x Editor s note Earlier interviews and Q amp A sessions with Don can be found in these issues of TUGboat e 7 No 2 1986 The coming out party for Computers amp Typesetting May 1986 e 13 No 4 1992 Q amp A with the Nordic group and conversation with Ros
18. witha Graham November 1991 e 17 No 1 1996 TUG 95 Q amp A e 17 No 4 1996 Q amp A Amsterdam NTG and Prague CSTUG March 1996 e 21 No 2 2000 Interview by Advogato January 2000 Several of these sessions are already posted on the TUG web pages and the rest will be posted when time permits
Download Pdf Manuals
Related Search