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

At 15:10 29/03/97 -0800, Terry wrote:
>Let me say it another way.  User agents should not be required to
>support catalogues *or any other mechanism specified by XML*
>to resolve public identifiers.  There should be *no default
>assumed* by publishers.  That's hard to swallow, and most of 
>this group is choking on it; ask yourselves whether you need
>a specified default mechanism to find a copy of Moby Dick;
>ask yourselves whether, had Melville's publisher specified a
>default mechanism (e.g., writing to the publisher's address),
>you should have to follow it today.

This is similar to what I had in mind. There are enough bookstores in
existence that I can walk into pretty much any one of them and buy a
copy of M. Dick, and the author/title is enough because it's unique
(I don't _need_ the ISBN of a specific edition [read: URL] unless I
actually _want_ that specific edition and no other -- which some people

When FPI resolvers are as ubiquitous as bookstores, my XML/SGML app
will just hit the nearest one with an FPI and say "gimme" (and $0.02
cyberbucks will transfer ownership :-)  If it's a rare FPI, say an
early copy of Docbook :-) then the resolver might say "I can't do this,
try down the street at Harry's Rare Resolver, and failing that, try
the Public Owner's own site". This last will get a horse-laugh from the
DNS or whatever because HaL and O'Reilly may not may to Davenport...

The point is that the data structure (the FPI) already exists, and people
want to use it, even as a placeholder. Even if it's a year before any
widespread resolving mechanisms are in use, I see no reason to prevent
its use, and if the current network only resolves URLs then we should
document that use of URLs in sysids is mandatory for the moment.

A happy Hale-Bopp to you all.