Re: public/catalog [was: Re: ERB Decisions of March 26th]

At 00:41 31/3/97 -0500, Paul Prescod wrote:

>Similarly, providing a catalog provides you with a starting point that
>is at least as robust as a URL, and if you know a better way to find the
>public identifier, you are encouraged to use it. Michael's current
>proposed spec (which is newer than your message) is quite explicit about

Actually catalogs are more stable than URLs. Because a catalog is likely to
identify a large number of documents from a single URL it is probably less
likely to change its location than the individual URLs are. Its length can
also vary over time to make it more extensive. This is particularly
important where catalogs are used to indicate sets of related documents,
where the catalog can also be a guide to the set of currently available
documents on a subject.

>So your analogies of forcing booksellers to write letters to the
>publisher of Moby Dick are inaccurate.

The bookshop analogy is, in fact, not the most valid one. A much more valid
one is that of the Librarian wanting to correctly locate a book like Moby
Dick. You don't catalog such books by giving them an absolute address like
"the third shelf of the fourth stack" as this changes too frequently. You
simply classify books by category and then list them in alphabetical order
in that category. In terms of what we are proposing you would used one
catalog per category, and a public identifier to provide a sortable list of
unique names for the book e.g. "-//library1234//TEXT Melville/Moby Dick//EN".

You need public identifiers to overcome the location shifts so prevelant in
modern nucleated systems. Without public identifiers, and catalogs to
intepret them, you cannot create any valid WWW document set with a realistic
lifespan of more than 2-3 years as that is the average lifespan of the
underlying system, and a new system typically means a revised file
structure. With public identifiers, and an ever expanding set of resolution
methods, you can envisage this mechanism being expanded to 20-30 years (but
not beyond, because by then no-one will be using those arcane DOS-based URL
file paths to locate things beyond about 2010: we will all be using URNs and
other forms of public identifiers, such as EAN bar codes and numeric
Martin Bryan, The SGML Centre, Churchdown, Glos. GL3 2PU, UK 
Phone/Fax: +44 1452 714029   WWW home page: http://www.sgml.u-net.com/