Re: TAG request: establish the relationship between URIs and Resources is many to many

I agree with what Tim writes, but I think confusion will be reduced if we 
can drill a bit deeper on the following common example.  Consider:

        http://example.org/mydoc.version1
        http://example.org/mydoc.version2
        http://example.org/mydoc.version3
        http://example.org/mydoc.latest

...where as a matter of policy I cause GETs to "latest" to return the same 
representations as references to the version with the highest number.  If 
I introduce a "version4", then GETs to "latest" will start tracking that. 
I know this is not a new use case, but I think we could do a better job of 
dealing with it in the latest round of discussions.

In some simple sense, this can be viewed as a situation where two URIs 
identify the same resource, with the resource identified by "latest" 
changing over time.  There is another sense in which "latest" is an n'th + 
first resource.    Here are two use cases that motivate these two 
seemingly conflicting views.

Some use cases:

* RDF should presumably be able to make separate statements about version 
3 and about latest (e.g. "the version you should read is the one 
identified by "http://example.org/mydoc.latest", 
"http://example.org/mydoc.version3 has a bug").  So, there is an important 
sense in which there are two "things", which perhaps should be called 
resources.

* If I do an HTTP POST or PUT to http://example.org/mydoc.latest, the 
representations returned by GET's to version 3 (or 4) necessarily change. 
So, in this sense we can see that there is one "thing" under there.  Do we 
care to model this possibility formally, or do we view the correspondence 
accidental?  Stated another way, do we wish to say "the two URIs refer to 
the same resource (but http://example.org/mydoc.latest may refer to a 
different resource tomorrow)" or do we prefer to say "by definition two 
different URIs refer to two different resources...the fact that we observe 
a correlation between their representations is not modeled at this level 
and is not directly known to RDF"?

I'm not promoting a preferred approach to clarifying the handling of this 
case, but I do think clarification is necessary.   I'm not sure that the 
latest proposals provide unambiguous answers for these scenarios regarding 
correct use of the term "resource" or for interpreting statements like 
"The Web is designed on the principle that a single URI identifies a 
single resource which does not change."  I would find it helpful to see 
the various proposals worked through in the context of these use cases.

Thank you.  I hope this helps drive the analysis in a useful direction.

------------------------------------------------------------------
Noah Mendelsohn                              Voice: 1-617-693-4036
IBM Corporation                                Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------

Received on Friday, 31 January 2003 12:07:45 UTC