Stephen Frost wrote:
> I don't think either message really fits here, unfortunately. We're not
> actually checking the uniqueness of someone else's tuple here either,
> after all, we're waiting to see what happens with their tuple because
> ours won't be unique if it goes in with that other tuple in place, if
> I'm following along correctly.
FWIW I think the translatability argument carries very little weight.
We add new messages to the catalog in minor releases every now and then.
We don't take it lightly of course, but avoiding so isn't a reason to
not fix a bug properly.
In this case, given the discussion, it seems to me that the right fix is
to create a new XLTW enum as I already mentioned, with a message
tailored to the specific case.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services