The Jargon File (http://www.science.uva.nl/~mes/jargon/t/top.html)
reports that in 1989, hacker Paul Boutin was asked ``What do you think
will be the biggest problem in computing in the 90s?'' Paul's laconic
answer was ``There are only 17,000 three-letter acronyms.'' (Those
hackers among us will immediately counter that, to be exact, there are
263 = 17,576.) Do you know that over half (http://www.atomiser.demon.co.uk/abbrev/) of the
possible TLAs (three-letter acronyms) have been allocated?! I wish to
talk about a particular TLA: LPU (and I don't mean the airport code of
Longapung, Indonesia).
I've been corresponding with an august group of over a dozen
conference and program chairs, trying to solidify the conference
schedule for 2002. It turns out to be very difficult to meet the
following three objectives we set for ourselves: (1) the notification
date for one conference should precede the submission date of the next
conference by at least three weeks, to allow for adequate revising,
(2) the notification date should follow the submission date by at
least ten weeks, to allow for adequate reviewing, and (3) no deadline
(submission, notification, camera-ready copy) should be simultaneous
with another deadline, with a conference, or with a major holiday. We
think we've finally gotten it, but it has taken several iterations,
and flexibility on the part of everyone.
This negotiation brings to mind the commonly observed phenomenon of
repeatedly submitting a paper to conference after conference until it
gets accepted. Those who have served on multiple program committees
quickly notice papers going through three, four, even five conferences
before being accepted, or, much more frequently, the author finally
giving up. And our effort to arrange for non-overlapping reviewing
periods, where there is the opportunity for rejected papers to be
submitted to the next conference, may in fact encourage this behavior.
Why are there so many poor papers? I feel that one cause is a
prominent paper-writing strategy called ``slicing and dicing''. The
objective of this exercise is to get the most papers from the research
results of one's program of study. If the papers are cut too thin,
they don't get accepted. If too much is put in each paper, fewer
papers result. The solution is to identify the LPU (Least
Publishable Unit): that kernel of results that is broad enough
to just barely be accepted by a conference or journal. If one is
lucky, a research program might have within it four or five LPU's,
which will look very good on one's vita.
You can peruse most any conference proceedings and identify papers
written with this strategy. It is also possible to compare the LPU
granularity of various conferences. An LPU, say, of SIGMOD, is easily
two or three LPU's in some other database conferences, not to be named
here.
The problem is that this strategy also yields papers that have the
lowest potential utility (another form of LPU). The world
is awash in write-only papers. As an indicator of this, the Anthology
(www.acm.org/sigmod/dblp/db/about/top.html) lists the
top 122 referenced papers, out of 100K citations that were analyzed.
These papers alone totaled 16% of the analyzed citations. If we
assume that every paper references on average 20 other papers, then
this portion of the skew alone relegates some 7000 papers to not being
referenced by a single other paper. A paper that is never referenced
is likely to have little or no impact. What is the use of many papers
on a vita if they are never read by anyone else?
I recommend the opposite: that authors try to achieve the MPU (
Maximal Publishable Unit, or Most Potential
Utility) with each paper. Try to write a paper that may be the
best in the chosen conference or journal. This may not
maximize one's vita (at least the quantity in one's vita), but will
help to maximize the impact.
However, life is not so simple (it never is). It is easy to arrive at
a paper that has too much content, which reduces the chance of
acceptance. My first conference submission was to SOSP, the top
operating system conference (yes, I started out as an O/S guy; it took
me a while to see the light). It was rejected, and in retrospect,
rightly so. I submitted it to TOCS, where it was
accepted. Quite simply, the story I was trying to convey with that
paper was not tellable in 20 pages, but was possible in 40 pages. So,
some results are not appropriate for conferences: they require too
much set-up, the proofs are too complicated, whatever. So one must
first determine whether a result is appropriate for a conference, for
a journal, or even for a book (the story of my last book could not
have been told even in a journal paper). Then, if the paper is
destined for conference, determine the conference for which the paper
represents that venue's MPU, and submit it there. The chances are
quite good that the paper will be accepted. Ditto for journals. Ditto
for book publishers. There's my unsolicited advice, free for the
taking.
Speaking of conferences, I hope that you have registered for the SIGMOD'01 conference, to be held in Santa Barbara on
May 21 - 24, with the co-convened PODS'01 conference and several
workshops. When you come, don't forget to bring a spare database
textbook to donate, helping disseminate the ideas our community has
originated and nurtured. This year, books will be donated to Egypt.
I want to thank Dave Maier, who rotates off the SIGMOD Awards
Committee. Dave has served three years on this committee, the last as
chair, always providing cogent input and advice. I thank him for his
efforts, and also thank Phil Bernstein, who has taken over as chair,
and David DeWitt and Jeff Ullman, who join as new members, for
agreeing to serve on this important committee. The perceived worth of
the SIGMOD awards is directly correlated with the stature of those on
the committee, and this year's committee lends great prestige to the
awards it names.
The SIGMOD Elections Committee, under the able direction of Meral
Ozsoyoglu, has assembled an impressive slate of candidates. It is
such a comfort to know that SIGMOD will be in good hands. My term ends
in June, and so this is my last chair's message.
My tenure as chair has been truly a wonderful, incredible
experience. SIGMOD is full of amazing people, willing to invest
generous amounts of time, skill, and passion in activities that
benefit others in our community. The first page of this SIGMOD
Record lists many of these people, but there are also many more
working with the conference and other activities. My joy over the past
four years has been to interact with such people, and to get to know
them as colleagues and as friends. I am deeply grateful for the
opportunity you have given me to serve you, and I thank you from the
bottom of my heart.
See you in Santa Barbara!
Rick Snodgrass |
January, 2001 |
|