[TUHS] How to Kill a Technical Conference (was: Zombified SCO comes back from the dead, brings trial back to life against IBM)

Theodore Ts'o tytso at mit.edu
Tue Apr 6 00:05:04 AEST 2021

On Sun, Apr 04, 2021 at 02:22:22PM -0400, Clem Cole wrote:
> Actually, quite the opposite, USENIX was getting more and more academic and
> research-oriented and less 'trade show.'  The key is that USENIX and ALS
> should have been an excellent match, unfortunately, some of the
> personalities involved were at odds with each other.  IMO: it was more of a
> crash of personalities/control issues - the details do not need to be
> repeated or aired again. Note: I was on the BOD at that time and in fact on
> the PC for that specific conference.  Ted may have been on the BOD at the
> same time.

A lot of the conflicts were over the writing style.  There many ways
you can write papers.  For example: 

   (a) academic papers suitable for tenure-track publications
   (b) technical industry paper meant for other industry practitioners
   (c) white papers written by and for sales/marketing folks

Usenix at the time was trying very hard to make sure its conferences
would be taken seriously by tenure committees, and so there was a
strong sense that any paper published by Usenix had to be super high
quality from the perspective of (a).

Most of the Linux programmers had absolutely no interest in getting
tenure, and with much prodding, you might be able to get them to
strive for (b).  But there were a lot of people who were volunteering
for Usenix program committees, even for things like USELINUX, who were
convinced that (a) was the only way to write quality papers.

There was a bit of, "well, we'll give you a kiddy-pool track called
USELINUX, but maybe someday you'll grow up and write real papers."
That condescension was probably responsible for a lot of the
personality issues.

There was a lot of other things going on around that time; which
others have written about or observed --- the passing (into
retirement, or being kicked up into senior management) of technical
leaders at companies who would encourage their engineers to publish at

IBM stopped paying financial incentives for people to publish papers
at conferences, and it became clear it didn't really help ones career
prospects, with the possible exception of once you were trying for DE;
another company with which I have a lot of personal experience
strongly discouraged people from publishing because there was a slow,
bureaucratic process required to get publication permission lest that
company's "secret sauce" get exposed, etc.  Interestingly, that
process only applies to *papers*; but if you give a *talk* at a Linux
conference, it was a lot easier to bypass the bureaucracy.

> But USENIX also published less formal papers.  In fact, one of my
> all-time favorite practitioner papers is from another member of this
> list -- Tom Lyon's "*All the Chips that Fit*" from the 1985 Summer
> USENIX [which if you have never read, send me an email, offline and
> I'll send you a scanned PDF -- note to Tom if you still have the
> original bits I bet USENIX would like them]. I suspect that such a
> paper would never have been acceptable in any of the IEEE or ACM
> conferences.

Yeah, that was the 80's.  By the time the 90's rolled around, people
were trying *really* hard to become much more formal.  And if you
compare papers from a conference like, say, FAST, I'd say they are at
the same level as many ACM/IEEE papers.  I will be the first to say
this is a really good thing, but it does have some engineering

> Back to your point, USENIX may have stopped being as important to
> many practitioners, particularly ones in the FOSS community. Which I
> do find sad, but I understand the issues on both sides and why that
> might be so.  For instance, Keith Packard of X11 fame, Steinhart,
> and I were all talking about "whence USENIX" at a Hackers
> conferences a few years back.  So, if you come from that side of the
> world, you may not value membership or the results (BTW: my own now
> hacker daughter, who is a Googler, dropped her membership last year
> as she felt it was of less value to her); but so far USENIX has
> continued to be important to a large part of the research community
> and a set of some practitioners.

Everyone has the best of intentions, but reality is that the
incentives for academic researchers and industry practitioners have
widened over time.  Everyone has a limited amount of time and budget,
and if you're an industry practitioner, you are primarily graded on
your ability for you to deliver technically difficult projects that
have great impact on the company --- and the easist ways for impact to
be judged by a promotion committee is by saving the company $XXX
million over 5 years, or bringing in $XXX million.  If you are an
academic, the deliverables by which you are judged are graduate
students and publications.  And if you don't have tenure yet, those
had better be publications that are taken seriously by tenure

So creating a conference which meets the primary goals of these two
populatations is... hard.  Quite frankly, it's a lot easier for me to
make a case to go to Usenix conferences to (a) harvest ideas for me to
take back to the company, and (b) harvest graduate students for the
hiring pipeline.  It's not to contribute to the conference; I might do
that on my own time, but realistically, when I spend 80-100 hours
working on a program commitee, that mostly comes out of my own
personal time --- not really out of time paid for by my company.  At
best my company will tolerate my time to attend a PC meeting, and pay
for my travel.  But that's really about it.

And so to spend a huge amount of time writing a paper which is
suitable for a program commitee which has been giving its marching
orders by the conference's steering committee to make sure that the
proceedings will be taken seriously by a tenure commitee?  Nope, not
the best use my time.


						- Ted

More information about the TUHS mailing list