“Questions and Answers With Professor Donald E. Knuth”, 1996-03 ():
Q: …Now that PostScript is becoming so widely used, do you think it is a good replacement for METAFONT—I mean, good enough? Right now, we can use TeX and PostScript…
Donald E. Knuth: The question is, is PostScript a good enough replacement for METAFONT?
I believe that the available PostScript fonts are quite excellent quality, even though they don’t use all of the refinements in METAFONT. They capture the artwork of top-quality designs. The multiple master fonts have only two or 3 parameters, while Computer Modern has more than 60 parameters; even with only two or 3 it’s still quite good. The Myriad and Minion fonts are excellent.
I’m working now with people at Adobe, so that we can more easily substitute their multiple master fonts for the fonts of public-domain TeX documents. The goal is to make the PDF files smaller. The Acrobat system has PDF files which are much larger—they’re 10× as big as
dvifiles, but if you didn’t have to download the fonts, they would only be 3× as large as thedvifiles. PDF formats allow us search commands and quite good electronic documents. So I’m trying to make it easier to substitute the multiple master fonts. They still aren’t quite general enough. I certainly like the quality there.Adobe’s font artists, like Carol Twombly and Robert Slimbach, are great; I was just an amateur. My designs as they now appear are good enough for me to use in my own books without embarrassment, but I wouldn’t mind using the other ones. Yes, I like very much the fonts that other designers are doing.
Asking an artist to become enough of a mathematician to understand how to write a font with 60 parameters is too much. Computer scientists understand parameters, the rest of the world doesn’t. Most people didn’t even know the word ‘parameters’ until 5 years ago—it’s still a mysterious word. To a computer person, the most natural thing when you’re automating something is to try to show how you would change your program according to different specifications. But this is not a natural concept to most people. Most people like to work from a given set of specifications and then answer that design problem. They don’t want to give an answer to all possible design specifications that they might be given and explain how they would vary their solution to each specification. To a computer scientist, on the other hand, it’s easy to understand this kind of correspondence between variation of parameters and variation of programs.
…Then in 1982 or 1981, when I was writing TeX82, I was able to use his experience and all the feedback he had from users, and I made the system that became WEB. There was a period of two weeks when we were trying different names for DOC and UNDOC, and the winners were TANGLE and WEAVE. At that time, we had about 25 people in our group that would meet every Friday. And we would play around with a whole bunch of ideas and this was the reason for most of the success of TeX and METAFONT.
…But for many people it [LaTeX] is a simpler system, and it automates many of the things that people feel naturally ought to be automated. For me, the things that it automates are largely things that I consider are a small percentage of my total work. It doesn’t bother me that I hand-tune my bibliography, but it bothers other people a lot. I can understand why a lot of people prefer their way of working.
Also, when you’re writing in a system like LaTeX you can more easily follow a discipline that makes it possible for other programs to find the structure of your document. If you work in plain TeX, you can be completely unstructured in your approach and you can defeat any possible process that would try to automatically extract bibliographic entries and such things from your document. If you restrict yourself to some kind of a basic structure, then other processes become possible. So that’s quite valuable. It allows translation into other structures, languages and so on.
But I use TeX for so many different purposes where it would be much harder to provide canned routines. LaTeX is at a higher level; it’s not easy to bend it to brand-new applications. Very often I find that, for the kind of things that I want to do, I wake up in the morning and I think of a project . . . or my wife comes to me and says, “Don, can you make the following for me?” So I create 10 lines of TeX macros and all of a sudden I have a new language specifically for that kind of a document. A lot of my electronic documents don’t look like they have any markup whatsoever.