Shibboleth Developers

["Scott Cantor" <>]

> Ok, I will see if I can find something that can help. The only problem I
> see is that maybe this is backend-related, so you will have to check for
> the behaviour of every (supported) database.

I know, I'm not optmistic that I can detect that error. I'd probably have to
blindly retry on all operations, which sucks.

I'm convinced from the Oracle docs that I need to study a bit more. Oracle
does in fact prevent dirty *writes* when you use the top iso level (at first
I didn't think so, per my other note, but that made no sense, so I read some
more), but I'd still lose that on read-committed.

I need to reassess the code and see what the possible errors are, in light
of the versioning bit I added to that API. I forgot how much of that was
motivated by the need to potentially run this on databases like MySQL that
only pretend at transactional integrity. With that feature in there now, I
might be able to just drop the level and be done with it like you suggested.
It's an implicit lock as it is.

Thanks,
-- Scott