[openstreetmap-website] Switch to TIMESTAMP WITH TIME ZONE for current_* tables (#375)

[Matt Amos]

> The 'with timezone' variant is the idiomatic way to fix timezone issues. It was created for this purpose and is supported all the way down to the wirelevel protocol. You can fix timezone issues another way, but that's IMHO just a hack.

My experience is exactly the opposite: timestamps are `time_t` and always UTC. "With timezone" is a hack which adds unnecessary complexity, and clients supporting implicit timezone conversion are the reason that these bugs exist in the first place.

> You're hoping to outsmart the postgres devs, it's rarely a good bet :p You can either fix the timezone issue once server-side following best-practice, or multiple times (once per client) using one of the solutions above.
>
> That being said, if the "alter table" downtime is unacceptable to the sysadmins, then that theoretical discussion is moot, and the practical decision to fix things client-side wins.

I'm not trying to outsmart anyone - I'm just trying to avoid unnecessary complication and leakage of presentation-layer concepts into the database. But sadly these appear to be mandated by the SQL standard :worried:. We have already fixed the server, but unfortunately the database client libraries are apparently still broken by design.




---
Reply to this email directly or view it on GitHub:
https://github.com/openstreetmap/openstreetmap-website/issues/375#issuecomment-21484335
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/rails-dev/attachments/20130724/c578c23d/attachment.html>