> Several reasons.
>
> 1) I would still have the read transaction open for the entire duration
> so I would still have a stuck OAT.

if you made the read transaction readonly, this woul not stick the OAT
Alan

> 2) I am uncertain how locking would be affected.
> 3) For reasons of convenience I read the row to be updated and then use
> RDB$KEY in the UPDATE statement to find the record for update. This
> won't work across multiple transactions.
>