Barry Lind <blind@xythos.com> writes:
> Of these two limitations the first is more significant since users do
> issue 'commit' statements directly sometimes, whereas users would likely
> never change the GUC parameter in their SQL.
Well, we could fix either or both of these in the planned protocol
change. It would be reasonable for the Z message to convey not only
the current transaction state (not in xact/in xact/in failed xact,
I believe are the three interesting states) but also the current
autocommit boolean. So assuming that visibility of state is not the
determining issue, which way do you actually like?
regards, tom lane