Report from DSC 2005 in Seattle
Several weeks ago, I attended the 'Directions in Statistical Computing 2005'
(or DSC2005 for short) conference in Seattle that was hosted by the Biostats
department of U of Washington. [1] This email is meant to provide a brief
report back to Debian.
I. Well what is this anyway?
DSC 2005 was the fourth DSC confernce following previous ones in 1999, 2001
and 2003 all of which were held in Vienna, Austria. These conferences are
very close to the 'R Core' group of upstream authors of GNU R [2], most of
whom are academic statisticians. The DSC conferences have been instrumental
for the growth of R. In fact, prior to DSC 1999, no meeting had ever taken
place to get most of the R Core people into one place at the same time.
However, the conference is open to other (typically related) statistical
computing projects such as the FLOSS projects BioConductor, Omegahat,
XLispStat or, in this case, the commercial S-Plus. S-Plus is the commercial
implementation of the S language. Having started with the original S license
from AT&T in the 1980s, it was the sole implementation of the S language
until R came along in 1993. S-Plus is distributed and provided by Insightful
who are based in Seattle -- so Insightful kindly covered parts of the
conference costs and one of the social events.
Yet at the same time, this is an R conference, and is designated as the
'scientific' one with the 'UseR' conferences filling the in-between years
starting with last years UseR2004 and continuing next summer with UseR2006.
II. So what happened?
The format was typical of your small academic conference: presentations,
sometimes plenary and sometimes with two parallel tracks, and a lot of
hallway chats. This year, as I recall, all of the talks were about R or
S-Plus.
One interesting aspect was a panel discussion of how R and S-Plus may
cooperate more, and how S-Plus could leverage some of the success of R. This
isn't the first time the topic has been raised as Insightful's CEO even came
to last year's UseR2004 conference. Insightful sees that the research
momentum is more and more in the R camp as evidenced by the rapid growth of
e.g. the contributed R modules at CRAN [3] which now has over 600 packages
[4]. No clear solution emerged but the intent is to see what could be done,
but most people on both sides are at least in principle willing to work
towards more interoperability. Pesky details remain, licenses are of course
among them.
III. What does Debian have to do with it?
A few things, actually. Doug Bates, who originally packaged R for Debian and
who is co-maintainer with me, is actually part of R Core. Debian was by far
the first distro ship with R. Significant parts of the CRAN infrastructure
run Debian, as do the development machines of several R Core authors. (That
said, other Linux distros such as RH are also used quite a bit. as is OS X.)
Hence, among authors, developers and users of R, Debian is reasonably well
known.
I have presented at two prior conferences (DSC2003, UseR2004) about Debian
and/or my Quantian derivative (as I did this year) and am somewhat known for
covering R for Debian. I mentioned again that we plan to eventually have
build scripts to make all (of the currently already 600+) packages at CRAN
"apt-get'able" in some way or form. (However, the mini-project working on
this [4] is a little stalled, volunteers would be welcome). This is actually
easier than it may sound as the internal R package format bears some
resemblence to Debian package in terms of meta-data and the way which builds
and regression checks operate.
In sum, this was an excellent to stay in touch with upstream, learning what
upstream is planning and providing feedback from our packaging side. As
aside, Insightful also asked me to be part of an R/S-Plus panel at their user
conference in Princeton next week. Sadly, work obligation prevent me from
attending.
Finally, I gratefully acknowledge Debian's sponsorship of my Southwest
airfare to the conference thanks to prior approval by the DPL. This covered
around one third of my total cost.
Regards, Dirk
[1] http://depts.washington.edu/dsc2005/
[2] http://www.r-project.org
[3] http://cran.r-project.org
[3] http://cran.r-project.org/src/contrib
[4] http://pkg-bioc.alioth.debian.org
--
Statistics: The (futile) attempt to offer certainty about uncertainty.
-- Roger Koenker, 'Dictionary of Received Ideas of Statistics'
Reply to: