> No, I cache with autocommit OFF. Basically, I have code like this.
> Let me know if you see anything garingly wrong:

Seems to be ok, but I did not check it in details :)

My main point is that blob is represented by 8 byte ID. This ID is
stored in the result set row and is cached in the statement object.
When you access blob content, completely different API is used (namely
isc_get_segment and isc_put_segment function). In case when autocommit
is ON, we have to cache blob content and cleaning the references makes
sense. However, in case when autocommit is OFF, it makes little sense
to clear the reference to 8 bytes, you will not save too much.

Another question: would you mind to try new pooling classes from the
CVS version? We have added statement pooling there too and it would be
nice to get some feedback from you.

Roman