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

Re: WD for Namespaces 1.1

• To: <xml-dev@l...>
• Subject: Re: WD for Namespaces 1.1
• From: "Rick Jelliffe" <ricko@a...>
• Date: Wed, 10 Apr 2002 00:15:25 +1000
• References: <2C61CCE8A870D211A523080009B94E4306FEEB04@HQ5>

From: "Bullard, Claude L (Len)" <clbullar@i...>

> Namespaces are system flags to XML processors. 
> They don't belong in the core.  Because names 
> of resources and names of locations of resources 
> in the system are conflated, you can't put them 
> in the core without tieing them directly to 
> the system.  That violated platform independence.

I don't see that this is any worse than using non-formal
public identifiers in SGML:

 <!ENTITY xxx  PUBLIC
    "http://www.topologi.com/"
    "http://www.topologi.com" >

The point with XML is that it is not platform independent.
It is designed for the WWW. If anyone needs archiving
or platform independence for tricky jobs, they may
still have to resort to full SGML.

It is interesting to compare namespaces and the IS 8879
SGML idea of "Element Sets" (a group of declarations with
only ELEMENT, ATTLIST and I think NOTATION).

IMHO, Namespaces are just an extension of Element Sets
(*not* architectures, Steve :-).  Namespaces just make first-class
what is a second-class structure (declaration groupings) 
in IS 8879 SGML.    

That there needs to be RDDL interposed
between the namespace identifier and any actual schema
is merely that we need to support plurality; one can imagine
a RDDL server where referencing a namespace URI
with a RDDL purpose as a parameter retrieves a schema
directly.  

However the trouble with namespaces and DTDs is not
so much that one has to use Michael McQueen's PE 
technique for prefixes, which is tiresome; rather it
is because namespace were designed to allow
modularity, and highly-closed content models such
as supplied by DTDs (where ANY is the only way
to get openness) restrict modularity to modules known
at the time the schema was made.  

XML Schemas, RELAX NG and Examplotron are better in this 
regard, allowing various kinds of openness if you have the 
foresight to declare you element sets that way. Schematron
is neutral w.r.t. openness/closeness: it starts being open but
you add constraints; these constraints may themselves be
open (such as "A should contain a B") or closed ( such as
"A should contain only Bs").

Cheers
Rick Jelliffe

• Follow-Ups:
  • Re: WD for Namespaces 1.1
    • From: Tim Bray <tbray@t...>

• References:
  • RE: WD for Namespaces 1.1
    • From: "Bullard, Claude L (Len)" <clbullar@i...>

• Prev by Date: RE: WD for Namespaces 1.1
• Next by Date: Re: content model question
• Previous by thread: Re: WD for Namespaces 1.1
• Next by thread: Re: WD for Namespaces 1.1
• Index(es):
  • Date
  • Thread