[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: DNS based URIs that don't imply access method

• To: David Hunter <dhunter@v...>
• Subject: Re: DNS based URIs that don't imply access method
• From: Tim Bray <tbray@t...>
• Date: Thu, 25 Jul 2002 09:37:30 -0800
• Cc: "'xml-dev@l... '" <xml-dev@l...>
• References: <7AB2229D9557BF4AB41F2BB515D2C3DA37D9CB@torex2>
• User-agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X; en-US; rv:1.1b) Gecko/20020722

David Hunter wrote:
> At this point, I feel I need to intrude into the thread and ask [again, I'm
> afraid] for the benefits of using URIs for namespace names?

Well, the benefits of being able to piggyback on the DNS so that 
everyone can make their own and dodge collisions, as you note, are 
pretty large.

There's also the desirable side-effect that if you use an HTTP URI, you 
can put something there to help explain your shiny brand-new namespace 
to people and software.

There are two good arguments against namespace names the way they're 
done, one technical and one psychological.  The technical one is that 
the scoping of the declarations makes them rather awkward to handle when 
you're manipulating in-memory data structures that represent XML 
documents.  My personal way to deal with this is to use XML for 
interchange and mash the data into in-memory data structures that are 
optimized for application needs not XML fidelity, but there's no denying 
this causes problems for some people.

The psychological problem is the one you put your finger on, and the one 
that drives SimonStL to distraction: when you explain that these names 
are also URLs and are also names, only they're really URIs which are 
really in practice URLs only maybe more, only they're really URI 
*references* except for it's bad practice to use relative ones... well, 
you get some angst.

That granted, when actual programmers sit down to write actual code that 
deals with actual namespace-labeled markup, the implementions are 
straightforward and interoperate well, I have yet to hear a *single* 
story of an actual implementation problem caused by the URIness of 
namespace names.

If we had a blank slate and a chance to re-invent namespaces scratch, 
there are a few things I'd change (unprefixed attributes, undeclaration, 
no relative and no fragment), but I'd still use URIs.  -Tim

• Follow-Ups:
  • Re: DNS based URIs that don't imply access method
    • From: "Simon St.Laurent" <simonstl@s...>

• References:
  • RE: DNS based URIs that don't imply access method
    • From: David Hunter <dhunter@v...>

• Prev by Date: RE: DNS based URIs that don't imply access method
• Next by Date: RE: DNS based URIs that don't imply access method
• Previous by thread: Re: DNS based URIs that don't imply access method
• Next by thread: Re: DNS based URIs that don't imply access method
• Index(es):
  • Date
  • Thread