Re: FYI -- draft ietf uri doc

On Sun, Sep 10, 2000 at 12:03:37PM -0400, Al Gilman wrote:
> At 06:13 PM 2000-09-08 -0400, Leslie Daigle wrote:
> >Having sat back and watched for a while... there are at least 3 valid 
> >threads that I see:
> >
> >	. is there a general statement that can be made about
> >	  what a URI refers to (like the 'C' pointer confusion) --
> >	  is it the box, the contents of the box, the location 
> >	  where the box is if it in fact exists?
> >
> >	. are there useful mechanisms for expressing relationships
> >	  between resources, or between identifiers?  (a really
> >	  crude attempt at distilling Dan Laliberte's thoughtful 
> >	  comments).
> >
> >	. irrespective of whether the resource is the box, its
> >	  contents, or whatever, what are the services that
> >	  can make use of the identifier/work on the resource
> >	  (beyond the traditional, bald, "get").
> >
> >I think these are all valid and important issues in need of discussion.
> >But, at least the last 2 are outside the realm of the URI
> >specification itself (as it stands today, and many of us believe
> >it's best handled outside the identifier spec), 
> 
> Many of us believe it's best handled outside the identifier spec and some
> of us have been told it has to be solved outside the XML spec.  Therein
> lies a genuine problem. Until there is a better mutual agreement on how,
> end-to-end, it can and should be solved, the URI community should be
> supporting the investigation of avenues of relief.  We can't just blow this
> off.

Sure. But neither can we force the non-XML users of the URI specs to 
have to assume XMLs semantics and solutions. Yes XML is a rather important
piece of technology but it isn't the only consumer of these specifications.


> The record of the xml-uri@w3.org discussions
> <http://lists.w3.org/Archives/Public/xml-uri/> demonstrates, as I see it,
> that there is an integration problem between XML and URIs.

Yep. And IMNSHO, its XML that needs to be fixed, not URIs...

> The example that I would offer to suggest that there is an actual problem
> is the following:
> 
> How to reference a schema or other higher-level language specification from
> an XML document instance is an open issue.
> 
> There is a concept that inter-document references shall be by URI and any
> referential structure employing URIs shall accept any URI.  The 'URI' in
> "use a URI to identify the other resource" is an atomic notion, brooking no
> discrimination among sub-categories.

Which is one of the problems I had with the XML Namespaces spec. Heck, the
document itself attempted to do this but was either forbidden to or
did it poorly...

> On the other hand, the developers of XML processors feel it is only
> reasonable to expect that if processing an XML document is contingent on
> comprehending some externally referenced resource, that the XML processor
> has some assurance _from XML_ that it _can process_ that resource.

What do you see as the solutions? That this is problem for XML to fix or that 
the identifier infrastructure that XML chose to use needs changing?

> One could view it this way:  Hypertext grew up around the HTML <A> element
> used for informative references -- "for more information you may elect to
> recover..."  References to schemas where the correct processing of the
> current document instance is contingent on successful processing of the
> cited schema is another whole kettle of fish. It is a normative reference,
> as opposed to the informative html:a reference.

Yep....

> It is reasonable to consider that the terms and conditions of reference for
> normative references, where the referrer depends on the referred-to, should
> perhaps be more restrictive than the terms and conditions of reference for
> informative (see also) references.

Moste definitly. This is one of the larger issues in the whole XML Namespaces
debate of the past few years. One camp sees this whole infrastructure as
a bad thing and assumes that XML Namespaces are just syntactic tag
differentiation while another sees namespaces as the basis for future 
extension toward this 'normative reference' infrastructure.

> It is an open question as to whether this problem should be solved in the
> definition of a URI or if it should be solved through the ability to add
> normative assertions as to referenced type to X-Link references.  But if
> XML is indeed under a constraint to use "a URI, which has to be able to be
> any old URI" to refer to the schema the more restrictive terms of reference
> have to come from the URI and not the XML.  But that is not necessarily a
> given.  We have a problem.  We haven't got an agreement on an answser.

The solution I favored was to allow "any old URI" but be explicit about
what your using it for and why (basically Dan Laliberte's suggested solution).
Does that mean that XML 1.0 has issues? Yes. That's why we put version 
numbers on things. ;-)

There is a large (much larger than XML) installed base of software and
standards that depend on URIs having this freedom of choice. If we
attempt to change URIs to solve the problems of just _one_ of its consumers
then the pushback from the likes of LDAP, SMTP, ACAP, DAV, SIP, ENUM,
CALSCH, CNRP, BXXP, IMPP, FAX, IOTP, IMAP, etc is going to be
significant...

Nice assesement of the issues, BTW... 

-MM

-- 
--------------------------------------------------------------------------------
Michael Mealling	|      Vote Libertarian!       | www.rwhois.net/michael
Sr. Research Engineer   |   www.ga.lp.org/gwinnett     | ICQ#:         14198821
Network Solutions	|          www.lp.org          |  michaelm@netsol.com

Received on Sunday, 10 September 2000 13:04:03 UTC