Dave Walden: Please tell me about yourself.
Richard Palais: I was born and grew up in the area around
Boston. After completing both my undergraduate and graduate work at
Harvard in 1956, I spent the next two years as an Instructor in the
University of Chicago math department, followed by two years at the
Institute for Advanced Study in Princeton. Then in 1960 I returned to
Boston as a member of the Brandeis math department, and except for a
few years on sabbatical or leave of absence I remained there until
becoming emeritus in 1997. At that point I started a second career,
writing Mathematical Visualization software. The program I have been
involved with is called 3D-XplorMath. It is used quite widely for
research and teaching, and also for producing objects that are
incorporated into mathematically oriented art. It is freely available
at 3d-xplormath.org and we have set
up a virtual mathematical museum populated with objects produced by
the program at
VirtualMathMuseum.org. My wife,
Chuu-lian Terng,
is also a mathematician and was a
professor at Northeastern until about seven years ago, when she
received an offer of a position at the University of California at
Irvine, and we have now become happy Southern Californians.
DW: Please tell us about your involvement with and TUG.
RP: I became aware of a little before Don Knuth gave
his famous 1978 Gibbs Lecture in which he introduced and Metafont
to the world. But to explain how I became involved in the early
history of , and in particular how I became the first chair of TUG,
I will have to go back a number of years before that.
Like many U.S. academic mathematicians, starting early in my career I
became involved with the American Mathematical Society (AMS), first as
a journal editor and then as a member of its Council, the body that
sets its scientific policies. Later, in 1971, I was elected to the
Board of Trustees, the body that sets the budget and oversees the
rather extensive business activities of the AMS, and I served in that
capacity for the next ten years. Since the Board is rather small, and
I was the only one on it with extensive experience with computers, I
was assigned the oversight of the Society’s computer operations.
The decade of the 1970s was a grim one for mathematical publishing, or
perhaps I should say for mathematical typesetting, for that was the
source of the problems. Maybe in retrospect it can be seen as just a
period of difficult transition from the classic Monotype approach that
had served so well for so long, to the current golden age of , but
at the time the problems seemed almost unmanageable and we could not
at first see any light at the end of the tunnel. What happened was
that high-quality typesetting of mathematical manuscripts was becoming
more expensive and the quality of less expensive typesetting was not
acceptable. And since publishing mathematical articles, books and
reviews is the principal business of the AMS, the trustees faced a
continuing dilemma that was a constant headache: should we increase
sharply the prices of our products or should we give up on
high-quality typesetting. In the end we did both, and no one was
happy. In the less prestigious series we produced some pretty ugly
products, and where we felt we could not sacrifice our high-quality
standards we tried outsourcing to Korea and other countries where
costs were lower, but in the end we still had to raise prices.
I don’t pretend to be expert in the history of mathematical
typesetting, but from what I have read and heard, the following is
roughly the story behind the how and why mathematical typesetting went
into this downhill slide. In classic typesetting, with a Monotype
machine, the operator or compositor, can make a second pass over a
line of type and add new symbols at a different level. This makes it
possible for a highly trained Monotype operator to set a complex
mathematical display with subscripts and superscripts, and end up with
something that looks correct and even elegant to a mathematician.
However, with the slightly older Linotype system only a single pass
was possible, so while simple inline mathematical formulas could be
handled with Linotype, complex displays were out of reach. On the
other hand, the Linotype approach proved to be more efficient and
economical and Monotype gradually faded away, since the vast majority
of typesetting of books, magazines, and newspapers had no need for its
extra sophistication. By the late 1960s there were probably only a few
dozen Monotype compositors worldwide who were expert in setting
mathematics, and no more were being trained.
Monotype and Linotype originated in the nineteenth century and both
were “hot metal” processes; that is, they produced lines of type that
were cast from molten lead. In the 1960s a newer typesetting process
called photo-typesetting began to replace these hot metal machines. It
used photographic methods, projecting glyphs one by one onto
photographic film to produce the original printed page. While this
worked well for normal printing, it again proved less than
satisfactory for mathematics. In fact, it was dissatisfaction with a
proposed photo-typeset second edition of Volume 2 of The Art of
Computer Programming that Knuth says started him in 1976 on the road
to and Metafont. In the end the AMS did find an interim
photo-typesetting solution to its journal publishing needs from a
company called Science Typographers Inc. The software was a bit of a
kluge; it was difficult to use and the quality was less than stellar,
but we learned to live with it until came along and we could get
it up and running.
While I had heard rumors of Knuth’s interest in developing a digital
typesetting system somewhat earlier, it was only in late 1977, when I
saw the title of his Gibbs Lecture, that I became excited about it. It
so happened that I was the Chairman of the AMS Board of Trustees that
year, so at the special dinner for Knuth that preceded his talk I was
seated next to him. We went on to become quite good friends over the
next few years, but that was the first time I had met him and I recall
being struck by how friendly and easy to talk to he was. He seemed to
understand the difficulties we were up against trying to maintain the
high-quality of the typesetting in our journals and I had the
impression he might be willing to help us switch to for that
purpose. After I heard Knuth explain in more detail in his
lecture I became convinced that this was the way for us to go. But
perhaps the single most important AMS publication — its flagship so to
speak — was Mathematical Reviews which had a whole separate operation
and support staff in Ann Arbor, far from the AMS main office in
Providence, and anything as potentially disruptive to their work flow
as a complete change in the typesetting system had to be acceptable to
them. So when I found out that the MR people were as enthusiastic
about as I was, that settled the matter, and we were able to
convince the Executive Director and the rest of the trustees to make
the investment in . As it turned out, this was going to require
more than just money — it was also going to take the time and efforts
of a great many people to take full advantage of what had to
offer. For me in particular, it did absorb a lot of my time over the
next few years, but I felt well repaid since it also turned out to be
an adventure in which I had many new experiences and got to know many
interesting people.
Once we made the decision to go with , we were anxious to make the
transition as speedily as possible, and since the and Metafont
programs were written in SAIL, the Stanford Artificial Intelligence
Language, we installed in Providence a DECSystem 20, the system for
which the SAIL compiler was written. Most of the hard work required to
get up and running fell to Barbara Beeton. She went to Stanford on
the 1979 summer AMS- project, about which I will say more below.
One of Don’s graduate students, David Fuchs, taught her a lot about
nicalities, and she brought back an initial installation tape and
got working in Providence that Fall. Today we think of as the
epitome of a stable and unchanging program, but back then it was
evolving rapidly, and we needed a way to keep the AMS implementation
up-to-date. Fortunately Don Knuth had an account on an MIT computer,
the Macsyma development machine known as MIT-MC, and both it and the
SAIL computer were connected via the ARPANET, the precursor to today’s
Internet. So I got an account on the MC machine and each time was
updated, Don would FTP the new version to my account there and send an
email to palais@mit-mc to let me know. Late that night I would drive
in to the MIT Computer Lab and copy the new version to a reel of tape,
and the next day drive down to Providence to give it to Barbara.
The MC machine had a rather unorthodox home-brew operating system and
I occasionally needed help figuring out how to get something done on
it. Fortunately there was a young guru (who I later learned had
written much of that operating system) who always seemed to be around
and I surmised probably slept there. His name was
Richard Stallman and
he was always happy to help. When Barbara Beeton was at Stanford,
David Fuchs turned her on to Emacs, and she sorely missed having it on
the Providence machine, so when I learned that Stallman was also the
author of Emacs, I got AMS in touch with him, and the next time I
drove down to Providence rms, as he was known, came along to do the
installation. On the way down I inquired how much he was getting as a
consulting fee from AMS, and he seemed genuinely confused at the idea
that he should get paid for such a thing. I tried to convince him that
he should get at least $100 for his efforts, but I never checked and I
suspect he really didn’t care enough to ask for payment! I wasn’t much
surprised when later I heard that he had been named a MacArthur Fellow
or that he had founded the Free Software Foundation.
We had much larger goals in mind for than just being a new
production typesetting system for AMS journals. Early on we had the
idea that it would be a major step forward if we could get
mathematicians to use to typeset their own papers. This was not
aimed at saving publishers the expense of typesetting manuscripts but
rather at saving the time and effort of mathematician authors! In
those days after much work by an author and technical typist to get a
correct manuscript, many errors were often added by the compositor and
there would be one or more time-consuming and often frustrating
up-and-backs trying to get an error-free version ready for
publication. While the Users Group (TUG) was founded to be a
general clearing house for all -related matters, it was getting
mathematicians willing and able to actually “ their own papers”
that I and many others felt would be one of its important functions.
Of course we realized that writing a paper starting from plain
would not appeal to most mathematicians, so we decided to write a
special macro package that we called AMS-, specifically designed
for research mathematicians, to make it as easy as possible for them
to write an article in . I applied to the NSF for a grant to cover
the expenses for a group of mathematicians and AMS staff from
Providence and Ann Arbor to spend several weeks at Stanford working
with Knuth and some of his graduate students, learning how to write a
macro package. I expected that I and my friend
Robert Morris
would do
most of the actual macro writing, but I also asked another good
friend,
Michael Spivak,
to join us with the idea that he would write a
manual aimed at mathematicians, teaching them how to use AMS-. Mike
had just had a very frustrating experience trying (and ultimately
failing) to write a book using the UNIX troff typesetting system. He
had told me that the documentation for troff was miserable, so I told
him here was his chance to show how it should be done. The manual that
came out of this was of course Mike’s famous and successful Joy of
. But the big surprise was that Mike, who had never done any
computer programming before, was inspired by Knuth’s lectures on macro
writing and turned out to be a whiz at both designing a macro package
and writing macros, so he not only wrote the manual, but took over the
entire creation of AMS- from Bob Morris and me.
In Volume 1, Number 1 of the TUGboat, dated October 1980, I
wrote an article called “Message from the Chairman” in which I
talked about the role of the Users Group, and explained much of
what I said above about the goals we had in mind for . I added one
further, more futuristic sounding role for , and while we have not
yet completely achieved this vision, with the continual growth and
acceptance of JSTOR and the mathematics arXiv, we seem to well be on
our way there. Here is what I said:
If we go ahead a little further in time we can foresee a development
that I call the all-electronic, save-the-forest library. It is
rapidly becoming the case that many (perhaps most!) articles in a
particular library copy of a scientific journal are never read.
Those that are read are apt to be photocopied out of that library
copy to be read at leisure at home. The costs of printing, binding,
and mailing the journal makes up the other half of the journal’s
costs mentioned above. Why not save those costs too by having the
journal in magnetic storage at some (or several) central locations.
It will soon be cheap and easy to peruse such a magnetic journal on
a computer terminal in the comfort of one’s own home or office.
And then I noted that source files or DVI files would be a good
format in which to store such online journals. Of course today I would
have added PDF.
Shortly after writing that article I started to disengage from
TUG-related activities, and in particular Mike Spivak took over for me
as Chair of the TUG Steering Committee. There were several reasons for
this. For one, I wanted to concentrate more on my mathematical
research, and I spent the 1980–81 academic year on sabbatical leave
from Brandeis at the Mathematics Institute in Bonn, Germany. But
another reason was that I found myself inextricably involved in
various other interesting but time-consuming -related matters.
Perhaps I can best finish up by saying something about a couple of
those activities.
I mentioned earlier that the Trustees went along with the
recommendation to make the primary mathematical composition
system for the production of AMS journals. But there was considerable
uneasiness with this decision and even some opposition. For a core
activity of the Society to be dependent on public domain software that
was not commercially supported made people more than a little
nervous — particularly since that software was still a long way from stable
and AMS had no prior in-house experience with maintaining a new and
complex software system. It was decided that in order to build up the
experience and expertise required, the Society should set up a
committee, the AMS Standing Committee on Composition Technology, whose
charge it would be to oversee the transition from STI to . The
members of the committee included AMS staff from the Providence
headquarters and the Math Reviews office in Ann Arbor and also Don
Knuth and some of his graduate students and postdocs who were involved
with the project. I was asked to chair this committee, not because
I had experience or expertise in composition technology — when the
committee was formed I had neither — but rather to keep the other AMS
trustees informed on how things were progressing and to handle the
very substantial administrative details. I looked over my file folders
from the committee while preparing for this interview and I was amazed
at the volume of it all. By the way, it was also this committee that
set up TUG, which explains the close relation TUG had with AMS in its
early days. I felt strongly that an organized group of users and
perts would be an invaluable resource. In fact, I believe that for
any open source software project to prosper, it must have a
well-organized body of enthusiastic volunteers to support it.
I don’t recall who first floated the idea, perhaps Don Knuth, but the
Committee on Composition Technology decided that the AMS should
sponsor the development of a new mathematical font for use with .
The idea was that formulas set using this font should give the
impression of carefully handwritten mathematics. Originally we
proposed to call it the Einstein font, but later it was decided that
it should be named after a mathematician rather than a physicist, and
we eventually called it the Euler font and set up a Font Subcommittee
to carry out its design. In connection with his work on Metafont,
Knuth had become friendly with Hermann Zapf, one of the great type
designers of the twentieth century, and he was able to recruit him for
this project. The detailed design of the font was of course the work
of Knuth and Zapf, but I again agreed to chair this committee to
relieve them of the administrative details. I also recruited a few
senior mathematicians whose taste I trusted and together we examined
carefully all the symbols of the Euler font and gave our collective
criticisms. It was a lot of work, but it was also fun, and worth it to
be able to work with a great font artist. I’m not sure how widely
Euler has been adopted, but Knuth used it in the book Concrete
Mathematics that he co-authored with
Ron Graham and
Oren Patashnik,
and I feel that it gives the book a very distinguished appearance.
DW: Thank you very much for taking the time to tell your story.