Re: Personal view

At 11:18 AM 6/17/00 +0200, Daniel Veillard wrote:
> The simple string comparison has the advantage of simplicity,
>and should be slightly less costly (in time/memory requirements)
>but denies the fact that namespace name are URI references (and
>then possible anchors for associated namespace related metadata).

It sounds like the 'forbid' option you recommend effectively gives us back
that simplicity, at the cost of relative URIs.

>I also note that the current section 6. Conformance of Documents
>in the namespace REC nor 5.3 really says what an XML processor
>- in the general sense - should do with document not conforming to
>this spec, they still look well formed XML-1.0 stricly speaking,
>should the data be made available to the application with just
>a warning, or should this be considered an extension of a Well Formedness
>error - which is the case for redundant attributes at the XML-1.0
>level.

This has been a complaint for a long while; "doesn't have an Infoset" isn't
really a meaningful error like those created by the detailed list of
well-formedness and validity constraints in XML 1.0.

> My personal opinion is still that relative URI in namespaces should
>be deprecated, this will avoid a number of problems with existing and
>upcoming specifications. I would also suggest as a guideline in the
>revision of the namespace specification that steps 6 c) 6 d) 6 e) and
>6 f) of the algorithm specified in the section
>  5.2. Resolving Relative References to Absolute Form
>of the rfc2396 should be applied to URI expected to be used as namespace
>name (i.e. before namespace name creation, not when handling them). So
>that "http://www.example.org/./a" is still a valid namespace name,
>but be clearly flagged as "bad practice". I would also like to see
>the namespace spec extended to say what should happen to non
>conformant document, should this be considered like a WF error,
>or should a simple warning-like error be raised ?

I like the way you've expressed a shift of absolutization responsibility
from the processor to the author:
>(i.e. before namespace name creation, not when handling them)

I remain concerned that absolutization still needs some substantial cleanup
- especially in regard to some of the error conditions discussed on this
list - but absolutization at creation time avoids some of the stickier
issues, like non-existent base URIs.  

This approach brings 'forbid' closer to 'literal' in some nice ways, though
I'm still not positive that it avoids context-sensitivity (the file: and
private IP address range cases).  It seems like comparisons could be done
as literals if this comparison were adopted, which is helpful, though the
baggage of URIs (identification? retrieval?) is still very much with us.

Simon St.Laurent
XML Elements of Style / XML: A Primer, 2nd Ed.
http://www.simonstl.com - XML essays and books

Received on Saturday, 17 June 2000 09:40:25 UTC