Re: RFC: Central document registry (long)

peter Murray-rust (
Fri, 3 Mar 1995 16:04:29 +0000 (GMT)

Thanks very much indeed Christoph,

This is certainly a discussion we need at the moment. It's a
problem - as you say - that the Internet is facing. Several ideas have
been tried and their success or failure depends to some extent on how
well people conform to the model.

As you say our course is fluid with a small amount of constraint.
The opposite end of the spectrum is where a group of people decide to
write a hypertextbook with carefully controlled namespace and quality.
The original C++ course is adopting this (Peter Mueller in Berlin) and the
idea is that there will be a de novo hypertext with links to other
resources. I agree with this but have also suggested that they adopt some
of the things that we have found to work well here such as the communal
construction of related material. The VSNS-BCD course (biocomputing) is
also taking the hypertextbook route and in this case they are requiring
peer-review for each chapter.

There are two independent themes here:
- peer review and quality control
- active of passive registration of documents.

In the first the quality control will be in the glossary, and, I believe,
in the very wide spread of disciplines in the course. I AM TRUSTING THAT
*GENTLY*! So far I have complete satisfaction in what has been posted.
When a controversial matter was raised it was dealt with through the
mailing lists.

The second probably depends on the style of course. In C++, for example,
most of us agree that a uniformity of coding style throughout the course
is very desirable. Some terms are very difficult - much debate on the C++
lists - and a central approach is probably valuable. In our own course
with the borders so fuzzy I welcome diversity.

The registration spectrum probably covers three categories:
- register only the top of each hypertree and let a robot (spider..)
search and index it.
- register each document with a central server, but do not insist on
namespace control
- require namespace control for documents and also add meta-data.

(namespace means labelling things with unique and meaningful names, like
vsns/pps/course/backbone/index.html. meta-data means adding to each
document <META...> fields about what is in the document, when it was
created, who is responsible, etc. Meta-data is not widely used on the
Web yet but it should be!)

Tools exist for all of these. I have heard good reports of HARVEST which
indexes hypertrees (e.g. on a campus), but I haven't used it. Many of
the search tools have registration facilities (e.g. ALIWEB, LYCOS).
However I think it's very important to register the meta data as well,
because otherwise the document has less meaning.

Remember also that some of the documents in out hypertree have an
independent existence outside the course. The most obvious of these is
the PDB, but it is also true of the hyperglossary which I envisage as
extending to many other subjects besides proteins

Discussion on this will be most welcome


Peter Murray-Rust, Glaxo Research & Dev. (; (BioMOO: PeterMR)
Birkbeck College,, CBMT/Daresbury,