[OAI-implementers] Looking for 2.0 server with completeListSize in resumptionToken
Mon, 8 Jul 2002 12:09:34 +0100
----- Original Message -----
From: "Alan Kent" <email@example.com>
> On Thu, Jul 04, 2002 at 05:39:51PM +0100, Tim Brody wrote:
> > 2) Versioning is an issue that hasn't been addressed by the OAI
> > have made the assumption that the top-level element of an OAI response
> > contains a namespace identifier which can be resolved to a version. I
> > table of XMLNS -> version (e.g.
> > 'http://www.openarchives.org/oai/1.0/oai_getrecord' => '1.0'), with some
> > capability to account for variability in implementation (e.g. convert to
> > lowercase, remove tailing /).
> The protocol has versioning addressed fine. My question was more that
> as a *developer* I was interested in finding out which sites were at
> which protocol levels. I think you were referring back to my question
> about the list of servers on the OAI site. My comment was more "hey,
> for developers, wouldn't it be useful if the sites showed what version
> they supported so I didn't have to manually try each one, one-by-one,
> to find a 2.0 server". No big deal.
The repository explorer (thanks Hussein!) lists repositories in versions
1.0,1.1, and 2.0. That list isn't checked (?) so some of those base URLs may
be updated/missing, but is a good starting point for developers.
Otherwise I'm working on producing an OAI 2.0 mirror of every OAI repository
there is ...
I would be interested to know why you think versioning has been addressed?
For two agents to talk together in an any-version world there has to be a
consistent method for getting the protocol version, regardless of the actual
version implemented. While the version is wrapped up in an OAI Identify
response it's likely to "jump around" within the XML between versions.
Assuming the element will always be "protocolVersion" a node search could be
performed, but there's nothing to stop a repository putting a
"protocolVersion" element inside a description.
Which is why I look at the top level element's namespace, which should be
more reliable than a protocolVersion anyway (certainly I've forgotten to
"roll-over" the version number before, but the XML won't validate unless the
namespace is correct).
All the best,