W3C home > Mailing lists > Public > public-semweb-lifesci@w3.org > February 2007

RE: [BioRDF] URI Resolution

From: Booth, David (HP Software - Boston) <dbooth@hp.com>
Date: Wed, 7 Feb 2007 15:50:04 -0500
Message-ID: <EBBD956B8A9002479B0C9CE9FE14A6C20207688B@tayexc19.americas.cpqcorp.net>
To: "Jonathan Rees" <jonathan.rees@gmail.com>, "Matthias Samwald" <samwald@gmx.at>
Cc: "public-semweb-lifesci" <public-semweb-lifesci@w3.org>

I haven't been able to follow this discussion entirely, so please
forgive me if I missed something that should have been evident.  In
general I like the idea of using an ontology to express URI resolution
information, but I am also partial to Xiaoshu's pleas for simplicity.
To my mind, the ideal would be a simple ontology that can be
straightforwardly implemented *without* the use of an OWL reasoner --
perhaps just using string substitution and the like.  However, I do not
yet have a clear enough understanding of the use cases and proposed
solutions to know whether that ideal would be feasible.  

> From: Jonathan Rees
> This leads me to think that the main relation we are looking for is
> one between an information resource (which may have multiple URI's)
> and a *string* that is a URL that will help us to get its
> representation.  Maybe this is obvious.

Actually, I do not see it that way.  It seems to me that the most
important use case is one that involves pattern-based transformations of
URIs as strings, rather than specifying relationships between individual
URI strings and resource locations, so that simple rules can specify
transformations that affect large numbers of URIs.  For example, if you
have old URIs that start with http://old/ , and the resources now reside
at corresponding locations starting with http://new/ , then you can
simply specify the pattern matching rule of transforming:

	http://old/(.*)$ to http://new/\1

This transformation is entirely in the space of treating URIs as
strings, and it only pertains to information *retrieval* -- not resource
*naming*.  The old names can still be used in RDF with their existing
semantics, and the application that is doing RDF reasoning using the old
URIs never has to know or care that the information was retrieved using
some other URIs.

In other words, even though the transformation of a particular URI
http://old/foo to http://new/foo can be interpreted as asserting a
semantic relationship between the string http://old/foo and the
information resource that http://new/foo names, it seems to me that most
applications that are using the old URIs would not and should not care
about that relationship.  It could be hidden in a proxy, for example.

Of course, a URI resolution proxy may care about such relationships if
there were additional assertions made about the resulting information
resources.  Is that what you had in mind?  And would these additional
assertions really need to be made about the individual information
resources?  Or would it be adequate for them simply to be associated
with the URI transformation rule?  I guess the use cases beyond URI
transformation are not yet crystal clear to me.

> 
> The www notion of 'resource' includes things that have different
> representations (bits) at different times, and things with operations
> other than GET, so while I agree with the goal of being precise about
> these things and avoiding content negotiation, we're in the
> unfortunate position of being without identifiers for particular
> representations (snapshots, versions).

Note that HTTP defines the ETag response-header field for distinguishing
between particular representations:
http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.19

> 
> I'm updating my draft to include multiple choices for solutions, with
> Alan's ontology as one among several, and will try to capture what
> you've said. In your view, how would one find information (or
> represent the information needed to find information) about a
> non-informationresource such as foaf:name?

I'm not sure what you are asking.  For the foaf:name property, you
should be able to follow-your-nose by deferencing the namespace.  Are
you talking about the need to find information about a specific person
with a foaf:mbox of mailto:dbooth@hp.com , for example?

David Booth
Received on Wednesday, 7 February 2007 20:50:19 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 18:00:46 GMT