RE: PURL and 303's

John,

> From: John McClure [mailto:jmcclure@hypergrove.com] 
> 
> I don't recall seeing a note about the relationship of the 
> "thing-described-by.org approach" with the PURL service 
> provided by the OCLC (see http://purl.oclc.org/).... Is there 
> a need for both? Or is the PURL service ultimately intended 
> to be passe?

They serve different needs, so no, the PURL service would not become
passe.  

A PURL forwards to a URL that is dynamically looked up in a database at
the PURL hosting location.  The owner of the PURL can thus change the
forwarding URL at will by updating the database.  On the other hand, the
"thing-described-by.org approach" *always* forwards to the *same* URL --
the URL that is in the query string of the original URL.  This allows
the forwarding URL to be determined by syntactic inspection of the
original URL, which permits the client to optimize away the extra
network access to thing-described-by.org.  With a PURL, this extra
access is unavoidable, because the forwarding URL could change at any
time.  So a PURL gives you flexibility that a thing-described-by.org URL
does not give you, at the cost of a network access that cannot be
optimized away.

It is possible to use a PURL within a thing-described-by.org URL, like
	http://thing-described-by.org?http://purl.org/foo
but there would be no real benefit in doing so -- it just makes the URL
longer.  It would make more sense to just base your URLs on a PURL
service that does 303 redirects (in conformance with the TAG's guidance)
instead of 302 redirects.  (purl.org currently does 302 redirects; this
is an unresolved issue.)

> . . . 
> [The PURL site states] that the 'future' solution is to use URNs for 
> (library) resources, so I'd also appreciate hearing about the 
> relationship of URNs to the "thing-described-by.org approach".

I don't know what their URN plans are, but I will say:

1. The "thing-described-by.org approach" was specifically designed to
stay within the space of the http scheme.

2. Personally, I'm a strong believer in the value of using http URIs as
identifiers.  The ability to use HTTP to locate useful information about
a URI is very convenient, very powerful and well proven.  (Of course,
the decision to make such information available is always up to the URI
owner.)  This also does NOT mean that I think all such http URIs should
be blindly dereferenced all the time.  On the contrary, for efficiency,
they should only be dereferenced when necessary.  This was one of the
motivations behind the thing-described-by.org approach: It permits an
extra network access to be reliably avoided.  

David Booth

Received on Thursday, 8 September 2005 16:16:54 UTC