RE: Proposed disposition of Stuart Williams' comments on Metadata in URI 31

Going back to the original note in this sub-thread [1], Marty Schleiff 

> I found Metadata in URI pretty interesting.

Good, thank you!

> While it seems pretty thorough in exploring
> metadata about a URI's intended resource, it
> doesn't seem to address metadata about the URI
> itself.

I don't really agree.  It's true that the examples generally talk about 
metadata that involves the resource as well, but I don't think the 
substantive conclusions make the distinction.  Consider what is arguably 
the key constraint in the finding:

"Constraint: Web software MUST NOT depend on the correctness of metadata 
inferred from a URI, except when the encoding of such metadata is 
documented by applicable standards and specifications."

I don't think that does distinguish and I don't think it needs to, between 
metadata about the resource and metadata about the URI.  So, while I think 
this email has kicked off a useful discussion, I am not inclined to revise 
the finding.  I suppose that to some small degree that's because we 
essentially pulled the "publish" trigger in June, and have been trying to 
finally wrap this since then, but more fundamentally I just don't buy that 
only resource metadata is addressed. 


> As the TAG is encouraging use of just a single
> scheme (i.e., http) for all identifiers,

I'm not sure that's quite the position of the TAG or most of its members, 
but it certainly is true that most of us believe that the http scheme 
should be used wherever reasonably possible, and that such reasonable uses 
include many of the cases for which alternate schemes are regularly 
promoted.  As one counterexample, the https scheme is regularly defended 
by many as appropriately distinct from the http scheme.

> it seems the TAG should also provide direction on
> how to convey processing instructions to relying
> applications.  The only suggestion I have seen,
> which I think came from a TAG member, and which I
> think is not very well thought out, was something
> like the following:
>     Don't do this:  <newScheme>://<stuff>
>     Do do this:     http://<newSchemeOrganization>.org/<stuff>

And so on. I think this is indeed an interesting and important discussion, 
but I think it's better aligned with the discussions we're tracking under 
the banner of issue URNsAndRegistries-50 [2].  There's also some overlap 
with schemeProtocols-49 [3], should we get back to that.

It's certainly true that in some respects all of these issues do involve 
certain aspects of metadata in URIs.  My view, however, is that the 
finding "The use of Metadata in URIs" is appropriately scoped as it 
stands.  It discusses the general guidelines for knowing when you may look 
at a URI, whether metadata can be encoded in the URI (e.g. to benefit 
orderly assignment of URIs to resource or to promote bulk documentation of 
assignment policies), but it does not except in passing in a few examples 
talk about the way specific parts of a URI might be used to convey 
particular sorts of metadata.  I don't think I would have wanted to go 
down into such things anyway, but I'm especially reluctant at this late 

So, I propose that the concerns Marty raises be considered in the context 
of our discussions of URNs and Registries (and I suggest that further 
correspondence be under a separate email "subject").  Thank you.



Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142

Received on Saturday, 30 September 2006 00:33:12 UTC