Chris Bowlby <chris@pgsql.com> writes:
> Would this not create the potention for a dead lock if transaction1 is
> never completed, and still active for an indefinate period of time?
If trans1 later waits (directly or indirectly) for trans2, we'll detect
the deadlock and abort one xact or the other to clear it. If trans1 is
idle because the client is asleep at the wheel, that could be
unfortunate, but it's not a deadlock.
regards, tom lane