Carp - Good answer and really explains things!

When we were having Snapshot too old problems the other day we found some code which contained commits accross fetches... Tut Tut we thought, made the rollback segments bigger and removed them.

Next day - Out of index tablespace...? Weird.
A bitmap index had grown from 8Mb to over 500Mb during the update of 300,000 records. So, back in the commits go since the app is coded to run pants without them (hence my cacheing question thread!).

I thought I was going mad, and couldn't work out how an index had grown so much, but what you're saying is that it recompresses after updating... Would that be right?