Hi,

> GC should never be inhibiteded in READ COMMITTED case.

At least that's how it should work :)

> Only TIP
> cleanup may need to be inhibited, but I doubt it for retaining
> transactions as they are marked as committed AFAIR. I may perform a
> closer look on this code if you want.

No, I will not do any changes in JayBird for a long time now, just a
little bit of support in the list.

> If what I wrote is wrong then it is a bug and should be threated as
> such. You can post post appropropriate item to Bug Tracker along
> with testcase.

I don't have it, but that's the outcome of few discussions in the
devel list and on the conference.

> In isc_tpb_concurrentcy modes GC is needed to be inhibited because
> snapshot is preserved across commit.

Yes. And this was probably main argument against implementing
auto-commit like in InterClient. JayBird should work nicely with all
transaction isolation levels.

> This had to be fixed in 1.5. Have you got a testcase for this
> problem?

No, that's the outcome from discussion in the devel list.

> Anyway, I do not think that clear implementation of auto-commit mode
> is important as this mode is usually useless for serious
> development.

Personally I have no time to invest into this issue. If somebody feels
that autocommit mode is of great importance for him, we would be very
grateful if he can contribute better implementation to the project.

Roman