[OAI-implementers] Results for various sites trying POST requests
Tue, 12 Feb 2002 12:18:47 +1100
On Mon, Feb 11, 2002 at 12:21:04PM -0000, Tim Brody wrote:
> What would be very interesting would be to correlate your results with how
> OAI-PMH has been implemented at the repository - whether the admin has used
> one of the available APIs, languages, and so on.
Maybe as part of OAI 2.0 the Identify command could have a standard
slot (if its not there already) identifying the toolkit implementation.
Sort of like User-Agent in HTTP or implementation in Z39.50. I certainly
noticed some consistent errors which I *assume* are toolkit implementation
> (It's a shame that many repositories seem to code their own implementation.
> Perhaps OAI should be more pro-active in promoting the development of, and
> use of standard APIs, including standard error messages)
I am in two minds here. I think a good thing about OAI is that it is
a relatively simple protocol. Wide adoptance I think is better based
on easy implementation and easy verification. SOAP for example has
benefited greatly from an interoperability testbed. People put up
test servers strictly for the purpose of interop testing. People
agree what the answers should be, and then fire their clients against
each other's servers to see if the results are correct. This may
be out of scope for OAI - it depends on the scale you want to achieve
For example, my personal current interest in OAI is not for digital
libraries. I see that it has potential for any site with metadata
to reduce their traffic from web crawlers.
I do not know what OAI 2.0 is going to be like. If it is XML request
and response packets (as distinct from HTTP variables and XML responses)
then you can easily come up with 'doing OAI with SOAP' documents.
It might not be the ideal way to do it, but SOAP supports the concept
of document/literal encoding where XML can be (almost!!) verbatim
dropped in wrapper SOAP elements. Basically there would be a mechanism
to get it more widely noticed.
> All the best,
Thanks for the above and all your other comments.
ps: I have done another pass over various sites using POST having fixed
bugs in my client. Things worked better. Full results are at
Summary of failures (remember, bugs may be my client!):
XML parse errors: aim25, anlc, cogprints, NSDL-DEV-CU, SUUB
POST not supported: cimi, HUBerlin, lacito, physdoc
Resumption related: ethnologue, hsss
Other: aisri, CPS, EKUTuebingen, ibiblio, in2p3, mit.etheses, ota,
thesis, tkn UDLAthesis, yea
But on the success side, I have managed to collect 381,000 metadata
records so far. I wonder what our next internet bill is going to be! :-)