On Sat, 18 Dec 2004, Tom Lane wrote:
> The really serious problem with this is the assumption that the server
> has the same set of available timezones (and spellings of same) as the
> client does.
I did some testing here and found of the 558 available timezones (in JDK
1.4) that 526 were accepted on 7.4 and 527 were accepted on 8.0, this is
on an unstable debian box, but it seemed like a pretty good percent.
> In 8.0 and later you can presume that the server is following the
> zic database's set of timezone names, but I dunno whether that has
> anything to do with the Java universe. Pre-8.0, you'd be foolish
> to assume much of anything.
>
> It's not quite too late to make TimeZone be one of the GUC_REPORT
> settings in 8.0, if that would help your problem.
>
Well this just kind of reverses the problem, now Java must have the same
set of timezones as the zic database, but I think making TimeZone a
GUC_REPORT is a good idea because we may want to go that route down the
line and now we certainly want to detect if someone has manually set the
TimeZone behind our backs.
Kris Jurka
