HARPER@finfun.bitnet ("Robert Harper ", Finland) (10/16/89)
Here is an entry from the CLASS-L. Since the original text was rather
large I have edited out the last part concerning the software. If you
are interested in obtaining more information then please
contact Fionn Murtagh at the address below.
----------------------- original --------------------------------
(Distribution via CLASS-L list)
Dear Colleague,
It is proposed to set up a directory of research software, relevant to
classification, multivariate data analysis and similar areas. Such
software will be available over network. In general it will be obtainable
through direct contact with the software-writer. Details follow below,
together with sample directory entries. Are there routines of yours which
could be included in this directory? In a few weeks, the directory will be
located on the SBCCVM List and the Netlib automatic distributor.
Fionn Murtagh
Earn/Bitnet: fionn@dgaeso51
fim@dgaipp1s
Span: esomc1::fionn
Internet: murtagh@scivax.stsci.edu
-------------------------------------------------------------------------
*-------------------------------------------------*
| Directory of Research Software for |
| Classification and Multivariate Data Analysis |
*-------------------------------------------------*
Contributions are solicited for details of programs and routines in the
field of clustering, multivariate data analysis, and related areas.
The objective is to facilitate communication between researchers who are
working in these areas; and to make the software byproducts of other work
both known and used in wider circles. Hence, the aim is both to have a
Directory which is of use to the user; and advantageous to the software
writer.
Although exceptions are not ruled out, it is envisaged to have stand-alone
routines available, which fulfil some well-defined purpose (e.g. carrying
out a certain type of constrained optimization; generating all combinatorial
structures of a particular type; etc.) Source code, in a common high-level
programming language, is very much encouraged. While not excluded, the use
of proprietary high-level command languages (e.g. SAS/IML) is not encouraged
on the grounds of limitations on implementation.
A fundamental aim of the Directory is that the Directory itself, and
communication between interested parties arising therefrom, be based on
electronic communication, using the widely-available academic and research
networks. The Directory listing the software will be regularly updated on
the CLASS-L list (maintained at the SBCCVM Bitnet Listserver, State University
of New York, by W.H.E. Day and F.J. Rohlf) and on the Netlib automatic
software distributor (maintained at Argonne National Laboratory, by J.J.
Dongarra and E. Grosse). For the foreseeable future, the Directory will
be in free-text format (allowing a copy of the Directory to be obtained,
and editor-based searches to be carried out). At a later date, it may become
necessary to use an appropriate database manager.
The information in the Directory should be of sufficient clarity to allow a
potential user to estimate well the effort required to implement the
software. From the point of view of the contributor of the software
information, genuine inquiries are sought rather than burdensome requests.
As a research service, it must be stressed that no guarantee or warrantee
will be given by the contributor of information. Similarly the Directory,
which will be kept as accurate as possible, should be clearly understood as
a communication service and no responsibility can be assumed for possible
misuse of information or software obtained via this information.
In line with the expectation expressed above that only genuine inquiries
for software be received by contributors of information to the Directory,
it is requested of all users that they clearly reference the originator
of the software or that they give relevant, associated bibliographic
reference(s).
In the first instance, please send information on software, which could
possibly come under the scope of this project, to:
Fionn Murtagh, at the following address: fionn@dgaeso51 on Bitnet/Earn.
Here are the categories of information which should be addressed for each
entry in the Directory.
-------------------------------------------------------------------------------
1. NAME/PURPOSE
Name of routine and extensive statement of purpose.
2. REFERENCE
Reference in literature.
3. ORIGINATOR/DATE
Software writer(s) and date(s) of writing.
4. CORRESPONDENT/E-MAIL ADDRESS/DATE
Who available from; electronic mail address. (Date of this information).
5. SIZE/HOW AVAILABLE
Approximate size of software in lines of code; and in bytes or blocks or
some other storage measurement. Automatically available over network
(anonymous ftp or otherwise)? Available via ordinary mail (e.g. magnetic
tape, together with format used; diskette, specifying PC, Macintosh,
density, etc.)?
6. LANGUAGE/ENVIRONMENT
Language (if possible, state exact compiler used); operating systems on
which the routine can run; operating systems and environments in which
the software has *actually* been used.
7. LIMITATIONS
Possible restrictions on usage (memory requirements; graphics hardware
requirements; machine dependence of random number generators; etc.)
8. DOCUMENTATION
Is the code well-documented? Is an accompanying "README" file available
to facilitate installing, linking, etc.? What accompanying hard-copy
documentation is available? Is this available in machine-readable form,
and if so in what format (ascii, LaTeX, troff, etc.)?
9. TESTING
Are one or more sample data sets available?
-------------------------------------------------------------------------------
[The procedures for automatically accessing the Directory will be summarized
here at a later date.]