I sent you 2 more emails today. Please read those as well (those two
need to be read in order).
John Rose wrote:
> I will try to set up a network with two computers, but this may take
me a couple of days since preparing for a trip to the States. Just to
make sure at your end, before final 2.85 release could you do the same
test with two Linux machines (one for OAI server and one for GLI)?
I was able to access only one Linux whose Greenstone server can be made
accessible to the outside world, such that OpenArchives' validator can
have access to it, for instance. Other Linux machines may be protected
by firewalls or something, my usual Linux can certainly not be made
accessible to outside of the department or university and failed
instantly because of this. Perhaps this is the sort of problem that you
encountered there as well when you were trying to connect from server to
client on the same machine?
Therefore I tried to set up the OAIServer on the Ubuntu Linux and
connect via the GLI from the same installation. I had a simple
collection of 2 PDF documents made available over the GS OAI Server and
GLI was able to download these just fine.
However, I did hit a terrible snag: when trying to validate the OAI
Server against the OpenArchives validator, to check it was accessible
(since the server on my usual Linux machine wasn't accessible and so GLI
couldn't connect to either), I found that on Linux it now fails a new
OAI server test.
Then I went onto my Windows OAI server to check that that at least was
fully compliant. It failed a totally different test. This was not at all
the case some weeks ago when Mariana Pichinini asked me about the OAI
GetRecord action and I confirmed for her that it worked to and
everything validated fine. At that stage I ran the validation test on
Windows once more to check that it exercised GetRecord also and never
encountered any failure.
It looks like they have added several more tests to the official
validator, including an additional one on the startup page. I looked at
the code just now, and Greenstone's OAIServer has no code to deal with
the point it failed on, so the older version of the validator could not
have been testing that aspect out before now. (And the tests were always
the same: it didn't ask different questions each time, but the same ones.)
I will now have to try to get it all to work again. The failure on the
windows may be more complicated to fix than that on the Linux.