CALL FOR DISCUSSION: comp.math and comp.stat

[W C Newell Jr]

This is a two-part proposal, involving the creation of two new groups:

    comp.math  and
    comp.stat

Because the charters of these two groups are very similar, I am presenting them
in a single article.  It is hoped that the discussion thread will be easier to
follow as a result.  A separate vote will be held for each group.

The Purpose of COMP.MATH:

To discuss suggestions and problems related to the use of commercial and public
domain math subroutine libraries, such as IMSL, NAG, PORT, LINPACK, EISPACK,
and others; also, to discuss the use of self-contained software packages for
numerical analysis and symbolic math, such as Mathematica.  Primary emphasis
will be on the selection and proper use of available routines, as well as
special considerations for porting codes across multiple platforms or using
non-Fortran programming languages.  Discussions on the implementation of
algorithms, including architectural considerations for vector and parallel
processors, will be considered germaine.  Theoretical discussions will continue
to be the province of sci.math, sci.math.num-analysis and sci.math.symbolic.
Benchmark codes are also germaine, although may by necessity involve
cross-posting with other comp.* groups.

This group will be unmoderated until such time as there is sufficient interest
in moderation to justify a further proposal.

The anticipated volume of comp.math is 10-20 articles/day, including the
cross-posting of existing public mailing lists.  Assuming the group is created
and sustains this level of interest, there will be followup proposals to open
subgroups, most likely

   comp.math.mathematica  and
   comp.math.benchmarks

I expect the vote on this group to be very high; I have placed a side bet on
the question of whether or not it will draw more than 1,000 yes votes.

The Purpose of COMP.STAT:

To discuss suggestions and problems related to the use of commercial and public
domain statistical analysis packages, such as SAS, SPSS-X, BMDP, S, MINITAB,
and others; also, to discuss the use of subroutine libraries, or subsets
thereof, implementing statistical functions.  Primary emphasis will be on the
selection and proper use of available procedures and routines, as well as
special considerations for using self-written procedures in conjunction with
existing packages.  Discussions on data handling methods, including techniques
for dealing with frequently-used government and commercial datasets (e.g. US
Census data) are welcome.  Discussions on the implementation of algorithms will
also be considered germaine.  Theoretical discussions of statistical methods
will continue to be the province of sci.math.stat.

This group will be unmoderated until such time as there is sufficient interest
in moderation to justify a further proposal.

The anticipated volume of comp.stat is 50-100 articles/day, including the
cross-posting of existing public mailing lists.  Assuming the group is created
and sustains this level of interest, there will be followup proposals to open
subgroups, most likely

    comp.stat.sas
    comp.stat.spss
    comp.stat.bmdp
    comp.stat.s

I expect the vote on this group to be very high; I have placed a side bet on
the question of whether or not it will draw more than 2,500 yes votes.

Common justification:

The time has come to incorporate support for major application software
packages into the Usenet namespace.  Each of the applications named above is
now available across multiple platforms, in most cases including various
flavors of UNIX.  Additional UNIX ports have been announced as in-progress by
the vendors.  The popularity of these packages is steadily increasing, both in
terms of the number of users and the range of disciplines they serve.

Public LISTSERV- and Internet-based mailing lists exist for several of these
packages, and the demand for peer consulting using these lists is straining
their manageability.  The SAS-L list alone generates 10-50 messages per day
and has at least 2,500 subscribers worldwide.  Many users are reluctant to
subscribe because of the inability to filter such a large volume of material
when it is distributed via e-mail.  The newsreaders' presentation model would
go a long way toward solving this problem.  In addition, message propagation
delays would be significantly reduced for academic sites on the Internet by
virtue of using NNTP, as opposed to BITNET's store-and-forward transport.

Many academic sites, including the UW, are now actively promoting use of the
newsreader as a general-purpose conferencing tool.  Our users are beginning to
feel comfortable with the notion of going to the newsreader for information
gathering or for second-line consulting support, even when their applications
are running in different environments.  We would like to encourage this trend
by adding more application-specific material to Usenet, and thereby utilize the
Internet's increased bandwidth to its fullest advantage.  It is recognized that
some UUCP sites may balk at this; in my opinion, however, the increases in
total volume, and hence the need for some sites to be more selective in their
choice of feeds, is inevitable.  The benefits to the Usenet community as a
whole are very much worth the nominal amount of extra work involved.

Administrative details:

The discussion period will last through Friday, December 15, at which time it
is expected that a call for votes will be posted.  The deadline for voting will
be sometime in mid-January, to accomodate the schedules of academic users who
will be on vacation during the holidays.  The votes for each group will be
administered at this site using aliased IDs; interim tallies will be kept
confidential, as other UW staff members may choose to lobby in favor of either
group.

I do not read news.groups, and I do not intend to participate in the "debate"
over these proposals.  Associates will be monitoring the discussion and will
provide me with a summary.  Questions on the proposed charter(s) which require
elaboration may be e-mailed to me, and I will post a followup, if necessary, to
news.groups.  Malicious flames are not welcome and will be forwarded to site
administrators.

It should be noted that the details for mailing list cross-posting have yet to
be worked out.  It is not my intention to repost digests of Usenet traffic
back to LISTSERV-based lists, but I will entertain any offers of assistance in
this regard should there prove to be demand after the groups are in operation.

This site will provide archives, as soon as I can find suitable software to do
the dirty work for me.

In summary:

These groups represent a major service enhancement to academic and research
users with access to Usenet.  They stand on their own merits.  In addition, it
is hoped that their inclusion will broaden the Usenet audience and further
enhance the quality of the information database overall.  I encourage everyone
to give them unqualified support.

Bill Newell
Systems Analyst, Applications Consulting Group
University of Washington
WCN at MAX.U.WASHINGTON.EDU