Subject Re: [ib-support] Deadlock and wait
Author Ann W. Harrison
At 08:38 PM 7/26/2001 -0400, Robert F. Tulloch wrote:
> > Commit retaining solves neither problem. It doesn't release resources
> > and, in the case of a consistent transaction, doesn't update the
> > transaction's view of data.
> It does with readcommited.

Actually, even with read committed, it doesn't release all resources and
it doesn't change the transactions view of data. Since my bias is that
anything that doesn't provide a consistent view of data is not a transaction,
read committed fails that metric too.

> What exactly then does this do?
> *
> * i s c _ r o l l b a c k _ r e t a i n i n g
> *
> **************************************
> *
> * Functional description
> * Abort a transaction but keep its environment valid
> *
> **************************************/

Aside from making steam come out of my ears? Undoes the transactions
changes without closing open cursors.


We have answers.