Re: identify XHTML DTD by URI, not by FPI

Russell Steven Shawn O'Connor wrote:
> 
> On Thu, 17 Feb 2000, Dan Connolly wrote:
> 
> > Again, I've seen that claim, but nothing behind it. Who finds
> > them useful? And for what uses? And cannot those uses be
> > served by a URI?
> 
> Wouldn't a FPI be usefull over a URL in the following sencerio.

No. I don't see how... in fact, a URI/URL works at least as well...

> The W3C pays for it's domain name w3.org to InterNIC.  InterNIC accepts
> it, but for some reason thinks they haven't.  InterNIC sells the
> ``unregistered'' domin.  w3.org is immediately bought you.  w3.org becomes
> a porno site.  Not nothing ... ahem ... validates, because a URL was used.

Everything that uses a cached copy of the DTD continues to validate,
because, as I said, the binding between Idtd and the corresponding
text is guaranteed not to change. You don't ever have to go back
to the net to see if it has changed. If something unforseen
happens and the net gives you different content in exchange for
that identifier, you're free to assume that the net is broken
and continue to use your cached copy.

That includes caches implemented using the SGML Open catalog mechanism.

> Admiditaly using an FPI would have the same problem, but with a URL,
> *every* document must be changed,

Why? If the catalog says

	SYSTEM "http://www.w3.org/.../html.dtd" "html.dtd"

then it doesn't matter if the net is broken or unavailable or whatever.
Validation continues to work.

> but for a FPI, only the catalogs need to
> be changed.

Changed to what? How does an FPI help? How would all these folks
with FPIs and catalogs find out about what to change to?

FPIs and catalogs take us back to the days of independently managed
/etc/hosts files. This is what DNS is all about. Yes, there are
some operational risks, but when it hits the fan, you can
fall back to local managemenet of your cache, your catalog, and
your /etc/hosts file. But meanwhile, when things are working,
you can just rely on the infrastructure to do all that for you.

> As you well know, this is a reasonably likely sencerio.

yes... http://www.w3.org/People/Connolly/9903webtechs-snafu

We eventually got webtechs.com back, by the way. But now
NSI is screwing up our billing again. Stay tuned...

-- 
Dan Connolly
http://www.w3.org/People/Connolly/

Received on Thursday, 17 February 2000 10:41:29 UTC