xml-dev - Re: Schema at XML namespace URI to change

[Rick JELLIFFE <ricko@geotempo.com>]

"Simon St.Laurent" wrote:
> 
> At 09:03 AM 10/22/00 -0700, Tim Bray wrote:
> >I think it's easy to misread what Henry's saying here.  The actual
> >"xml:" namespace URI is http://www.w3.org/XML/1998/namespace (see
> >http://www.w3.org/TR/REC-xml-names/#nsc-NSDeclared) if you deref that
> >you find a helpful human-readable document.  A couple of suggestions:
> 
> Henry Thompson et al. seem to be doing a bit more than that.  It appears
> that if you dereference the namespace URI + ".xsd", you consistently get a
> schema back.
> 
> For instance:
> http://www.w3.org/XML/1998/namespace
> and
> http://www.w3.org/XML/1998/namespace.xsd
> 
> To me, that's establishing a convention about XML processing and namespace
> URIs without creating a formal document describing said convention - not a
> good thing.  It may just be W3C practice. I'd suggest, however, that the
> 'W3C' outweighs the 'just' in the previous sentence.

If this idea proves useful, it may be good to document it. But then
there would
be the danger of premature standardization. Despite the strong opinions
of many people about namespace issues, I think most people also know
that 
we are still gathering data and figuring out what approaches actually
work.

I think this convention from W3C represents welcome progress in
sophistication
for the use of namespaces.

The main objections to namespace=schema are three, I think:
 1) without conventions, if the resource identified by a namespace URI
is
    a schema, that blocks of a more general mechanism
 2) namespaces are used for generic processing: they must not be tied
    to specific versions otherwise generic processing wil fail everytime
    there is a minor fix.
 3) we need a general way for resource discovery

The main proponents for namespace=schema say
 1) it has to be simple to work
 2) people's expectations is of some connection
 3) a namespace is a language (TBL)

The conventions that namespace URI plus a well-known extension is the
default location for a related resources using a well-known document
type
seems to meet the namespace!=schema points 1) and 2) well, and provides
an 80/20 on the third. It seems to meet the namespace=schema
peoples requirements 1), 2) and provides a hygenic way for people
who want namespace=language to play their game without messing the
sandpit.

And the versioning convention disconnects namespace=schema further,
because the effect of making the namespace URI point to the current
version is that it becomes clearer that it one should not rely on
it to signal variant properties of a name, but rather the invariant
ones.
This promotes the idea that namespace URIs are generic identifiers
rather
than schema-specific identifiers.

Cheers
Rick Jelliffe