xml-dev - Re: Question About Namespaces and DTDs

[Tim Bray <tbray@textuality.com>]

At 12:13 AM 25/07/00 -0400, Winchel 'Todd' Vincent, III wrote:
>In Michael Kay's book "XSLT Programmer's Reference" (2000 Wrox Press), page
>25, he writes " . . . serious use of Namespaces is virtually incompatible
>with serious use of Document Type Definitions . . . "
...

>Based on what St. Laurent wrote, it seems to me, at first blush, that this
>could be solved by requiring unique namespaces prefixes.  Am I missing
>something?  I realize the W3C is not going down this path.  

Yes, it could.  However, it's hard to dream up a way to design the
prefixes to avoid collisions, given the size and complexity of the web.
So the idea is to leverage off the DNS, which allows anyone to pick a name
with pretty-good confidence that it won't get duplicated.  There are a 
couple of ways you could think of using the DNS; java-package-name style,
e.g. org.w3.xml or URI-style, e.g. http://xml.w3.org.  Either way, the
names get pretty long and awkward, thus the idea of mapping them to
prefixes in order to keep documents moderately reasonable.  

At the time we cooked this stuff up, there were some real objections to
using URIs for this purpose, and as I recall they fell into 3 categories:

1. Public Identifiers were built for this purpose, use them
2. URNs were built for this purpose, use them
3. Architectural Forms were built for this purpose, use them

I can't remember anyone, at any stage of the process, arguing in favor
of using unique prefixes. -Tim