Hi Tom, > This mess was brought up in pgsql-bugs [1], but the solution > I propose here is invasive enough that I think it needs > discussion on -hackers. > > [...]
I tested and reviewed the patch. It fixes the originally reported bug and looks good to me. > The only other way I can envision to remove the round-trip hazard > is to stop using alphabetic abbreviations at all in timestamp > output, and use numeric GMT offsets regardless of datestyle. > I doubt that would make many people happy. It would certainly > break a bunch of our own regression tests, and I expect it would > break other people's applications too. I think we need to introduce alternative data types e.g. datetime / datetimetz pair if we want to do this. -- Best regards, Aleksander Alekseev
