Tom Lane writes:
> That seems excessive. There is a case for reporting autocommit (if
> we don't decide to remove it altogether, which is still an open
> argument). I have not heard any complaints suggesting that we need any
> others.
If we don't remove autocommit altogether on the server-side, then this
whole exercise is nearly pointless, because it still won't allow a
noninteractive libpq application to go in, do PQexec("command") and quit.
Each such application would still have to either set the autocommit state
explicitly first or, the new proposal, query the autocommit status and
then issue a COMMIT or not, as needed.
--
Peter Eisentraut peter_e@gmx.net