The First Free Software Conference
Intro and Disclaimer
This is a slightly edited writeup of my experiences at the Free
Software Conference that was originally intended for a friend who was
unable to attend the conference. Everything here is my opinion.
Official info about the conference can be found
here.
Friday
The youth hostel in Boston is cool, large and friendly and kind of run
down. There are a number of communal rooms, a nice kitchen, etc.
From there I walked across the river to the MIT campus, which is
pretty imposing - I don't think I'd have liked it as an undergrad, it
has an industrial feel. One cool thing about it is that you can
pretty much walk across campus without going outside of a building -
Lurking Horror isn't exaggerating much. On the east side of campus
across the street from the conference center was the MIT press
bookstore, which was awesome. Anyway, I got to the conference center
kind of early and they didn't need help setting up so I bummed around
looking for people to talk to. The first person I ran into was Eric
Raymond - he's a short, energetic guy in his middle thirties or so; we
talked about the Jargon File for a bit before some other people showed
up. Then he went off to talk to somebody and I talked to a guy from
Bell Labs for a while. He didn't seem worried about the labs
splitting up, and (jokingly?) said that his side was coming out on top
since they were losing the people responsible for the Korn Shell and
C++ all at once. A little later Phil Hughes, the editor of Linux
Journal, joined us. He's a cool, laid back type with long hair and a
beard - I got to know him fairly well before the end of the
conference. Around this time pretty much everybody else started
showing up, so there was a lot of milling around and peering at
nametags, since most people hadn't ever met most of the others in
person.
At the reception later on, I sat at a table with Michael Bushnell,
Roland McGrath, and some others that I don't remember. Both of them
are both pretty young - definitely below 30, which kind of surprised
me. mib is interesting; a little nerdy, but also kind of flamboyant
and talkative; he holds strong opinions about lots of things. Roland
is pretty quiet and harder to talk to, and doesn't seem to like small
talk a lot. For a while the two of them were in a discussion/argument
about the Hurd, which was interesting - they've implemented something
cool in the Hurd which stops a misbehaving process rather than killing
it and forcing a core dump since some information pertaining to the
process is lost when it is killed. While I'm talking about the Hurd I
might as well mention that I never met Miles Bader, even though he was
at the conference. Later on I talked for a while with Chris Demitrou,
formerly of the NetBSD core group, now (apparently) of the OpenBSD
group - I didn't ask about it since I didn't know it existed until
later on in the conference. It seems to be yet another *BSD splinter.
Linus (pronounced with a short 'i', just like Linux) was there, but I
didn't try to meet him since he was pretty much surrounded all
evening. In fact, he was attracting more attention than rms, which is
my preferred explanation for why, when rms walked by our table and mib
asked him how he was doing, he said "annoyed" - when rms walked off
mib and Roland exchanged shrugs. rms is a shortish guy with long
black hair, and a force of personality that makes him seem more like a
religious leader than a hacker. You really have to meet him sometime;
he's eerily calm and kind of stifles arguments because he's so self
assured that argument seems silly.
Things broke up a little before midnight when the hotel staff cleared
everybody out so they could start setting up for Saturday. While
waiting for a train I ran into Lisa Goldstein, who I hadn't heard of,
but apparently ran the business end of the FSF for a long time. She
seemed nice; when I asked her about rms she said he was great to be a
volunteer for but not to get on his payroll; she wouldn't elaborate on
that. The opinions that I formed of the free software community that
evening seemed to be supported during the following days - I felt a
little like an outsider since a lot of the people there knew a lot of
the others. However, nearly everybody I talked to was very friendly
and willing to talk about just about anything or anybody.
Saturday
I got to the conference center in time to snag some breakfast; I sat
down with people I didn't know - one of them had an FSF account that
had been wiped out by an intruder. I went to the Linux tutorial that
morning, which I wasn't terribly interested in since it didn't go into
technical details, but it was the only morning half-day session, and I
wanted to do the advanced emacs tutorial that afternoon. It was
really interesting anyway since I learned a lot about the Linux
Journal, because Phil Hughes ran the session. rms has apparently
tried to get him to refer to Linux system as "Linux based GNU
systems," and he refused on grounds that LJ is there to report, not to
proselytize. So, yes, there is some tension between the two camps,
and it was quite apparent at the conference. Lunch that day was cool;
I was at a table with Eric Raymond and a bunch of others. Meal
conversations tended to be quite interesting, and as often as not
didn't revolve around computers. Then the advanced emacs tutorial
started; rms pretty much adlibbed it - he went through some of the
strange but useful features that people who use emacs on a daily basis
probably wouldn't have encountered. Then he answered questions for a
while and talked about future plans for emacs. They include making it
easier for new users, integrating word-processor like features such as
better control over alignment, centering, and things like boldface,
underlining, italics, etc. He also wants to integrate some of the
xemacs display code (proportional fonts, inline graphics, etc), but
apparently the authors won't agree to GPL it. The only real surprise
to me was that he's going to replace the elisp interpreter with a
scheme interpreter. I hadn't known that Sussman and Abelson were on
the FSF's board of directors; that might explain why they're jumping
whole-hog into scheme. Later on I went out to dinner with Phil Hughes
and some other guys, to a seafood place that served some good beers,
and later on I went out for beer with a pretty big group including
Linus and his girlfriend (who was visiting the states with him), some
of the Hurd people, Tom Christiansen, and a bunch of others, to the
Cambridge Brewing Company, which was awesome - they had a great stout
and a barleywine that seemed to be very good, but I can't be sure
since I'd already had about three pints of stout. I know I was
impaired since I tried to order a pint of the barleywine before I
realized what I was saying - luckily for me it's illegal to serve so
large a quantity (or so the waitress claimed). Anyway, this was a
pretty wild and loud bar, so I talked only to the people close to me,
and I've forgotten who they were. Again, I didn't meet Linus since he
was mostly paying attention to his girlfriend and it would have been
rude to intrude; he also turned in fairly early in order to figure out
what he was going to say the next day. Tom Christiansen is
interesting, he's a little obnoxious but mostly cool. He makes a
living with Perl, consulting and teaching classes and stuff like that.
The free software crowd definitely seems to be oriented towards good
beers; I like that a lot. I took off around midnight since I didn't
want to miss the last train; the subway stations were packed with
Boston's college crowd doing the same thing.
Sunday
Linus gave the keynote speech in the morning. He's a pretty
interesting guy; sort if quiet and hesitant, but also funny. He
seemed pretty well adjusted to the fact the he's created a minor
industry, but emphasized that good timing was an important factor in
making it what it is. He appeared to be somewhat shaken about a
couple of things: first, some group has gotten a hacked Linux kernel
to pass the POSIX.1 validation tests, and he thought that eventually
the mainstream kernel would do so too. Second, the only obstacles to
getting Linux registered as a real UNIX are another set of tests and a
$25,000 fee that some group is apparently willing to pay. Things will
get interesting if that happens. He also talked about the various
Linux ports; Alpha and Sparc are pretty stable, with MIPS and PowerPC
progressing well, and the multiprocessor Linux for Pentium boards is
working but inefficient. There has been a major reworking of the
virtual memory architecture in version 1.3, improving performance a
lot. The code freeze for 2.0 is in effect. Well, it must be more of
a code slush since some pretty significant patches have come out since
the conference. Then came the rest of the talks; I'll only talk about
the more interesting ones, though they were generally of high quality.
A CS professor from New Mexico talked about using Linux for OS
courses. Some students of his have done interesting things, including
putting Linux on top of a little real-time kernel, which can preempt
the entire Linux kernel; Linux runs in the idle time of the real-time
executive. They had also added a lot of instrumentation code to the
kernel, to measure statistics about buffer cache efficiency and things
like that. Some people from OSF in Grenoble have gotten Linux to run
on top of Mach 3.0, on the Pentium and PowerPC - you may have seen
things about this on the net. I was under the impression that Mach 3
is inherently slow, but they claim that their system outperforms a
stock Linux kernel in all of the areas that they've spent time
optimizing, and it can potentially be faster everywhere. One of the
guys from Red Hat talked about their package system, which I've
subsequently found to be very nice - I picked up a Red Hat 2.1 CDROM
for free at the conference, and it's now running on three machines
here including the new P6 (and my machine, as of today). The last
talk was by Peter Deutsch, who now makes a living from ghostscript.
He talked about alternative licenses for free software - I'd imagine
that rms didn't like that one. At the end rms gave the other keynote
speech. He's a great orator; he talked about why he got into the GNU
business, and how it's doing. He seemed somewhat put out that Linux
has usurped the Hurd's position, but was fairly confident that the
Hurd would win out in the end. I don't agree, although I think it'll
have a significant following in the academic community, as a research
OS if nothing else. There were BOFs that evening; I went to one about
Perl, where there was in interesting discussion of a Perl compiler.
Tom Christiansen seemed to be against it since it would interfere with
source distribution since people could just make binaries and give
them out. The other one I went to was a demo of the OSF Linux - it
was pretty amazing to use Linux on Mach on a PowerPC; it even ran X.
Linus seemed to be bemused. Later I joined a group at the hotel bar;
it was kind of cheesy but served guinness. Finally I was sitting near
Linus - he's a pretty real guy. That phrase was heard a lot at the
conference, as in "have you met Larry Wall? Yeah, he's pretty real,"
as if to emphasize that these net personalities are humans too. I was
sitting across from Len Tower, who I had (embarrasingly) never heard
of. He's the other one besides Abelson and Sussman on the FSF board
of directors, and definitely comes from the old school of hackers.
He's totally laid back, and was willing to talk about the old days.
He said the hacker culture is slightly alive, in Sussman and Abelson's
research groups for example, but the good old days are over. He's
somewhat unwillingly migrated from hacking towards the administrative
end of the FSF, since somebody had to do it. At the conference there
was an interesting contrast between the old hackers, long haired and
laid back, and the new ones; still casually dressed but more
aggressive, and argumentative. Len had a paternal attitude towards
mib and the rest; he said Roland had been programming for the FSF
since he was a freshman in high school. He said mib is really
talented and likes to think he knows better than rms on Hurd design
issues, even though rms knows at least six operating systems inside
and out.
You know how it's possible to be surprised by pronunciations of things
you only see in text? Well, I was. All these guys pronounce things
like system calls, function names, and variables how they look,
regardless of derivation. For example, sprintf would be "sprint" as
in run fast, and "ef"; insn is "in sen", rather than instruction.
Weird. Len took credit for persuading rms to use rtl (register
transfer language) as gcc's internal representation - it's responsible
for most of gcc's portability. I guess rms wanted to get something
working and extend it later; that would have been a bad move. I was
curious about rtl since it was developed here (UVA, I mean) by the
prof I took computer architecture from last semester. I'm waiting for
a chance to take his compiler course; it's supposed to be pretty
awesome. Anyway, I sort of alternated talking with Len Tower on one
side, and Linus and Tom on the other; which was cool. I missed the
last train home so I had to walk across the river in below zero
weather - ouch.
Monday
The morning tutorial I went to was about 4.4 BSD internals, done by a
Harvard CS prof and one of her students, who ported NetBSD to the
Sparc20. It was interesting, since I'm fairly ignorant of BSD
internals. afternoon session was about writing machine descriptions
for gcc, taught by rms. It was awesome - he went through the whole
thing - the tutorial went a full two hours overtime. Towards the end
I started getting worried about missing my plane. The group attending
was pretty small, so there was room for questions. I don't remember
their names, but two of the guys there are working on the GNU Ada
translator. I'm much more impressed by gcc than I used to be - the
back end is truly amazing. Like I said, it went overtime, so
unfortunately pretty much everybody had cleared out by the time the
tutorial was over, so I headed directly for the airport; that turned
out to be a good idea since I had to use three subway lines to get
there. Then, when I got to Washington I discovered that my car had
been plowed in up to the license plate. Luckily the motel had a spare
snow shovel!
Oops, I left out the media lab. Well, I didn't see much. I realized
over lunch on Monday that I wasn't going to get chance to see it after
the conference, so I headed over there with only a few minutes to
spare before the gcc tutorial. I avoided the floor with Negraponte's
office since I'd heard that the staff there is unfriendly, and walked
around the third floor, mostly looking at displays and peeking into
windows. Actually, a lot of the walls are glass so I didn't really
have to peek. I wanted to walk into one of the rooms and ask if
somebody could show me around, but I had only a little time and I
didn't want to impose and then run off a few minutes later. I didn't
meet Vernor Vinge at the conference - he was on the program committee
but didn't make an appearance as far as I know. Furthermore I never
made it into the LCS, also due to time constraints - I really wanted
to see the FSF headquarters. Maybe next year...
Comments and criticism are welcome.
Last modified: Thu Feb 6 00:48:33 1997
jdr8d@cs.virginia.edu