Kevin Grittner wrote: > I just took a quick look at the spec to refresh my memory, and it > seems to require that the WITH TIME ZONE types store UTC (I suppose > for fast comparisons), it requires the time zone in the form of a > hour:minute offset to be stored with it, so you can determine the > local time from which it was derived. I concede that this is not > usually useful, and am glad we have a type that behaves as > timestamptz does; but occasionally a type that behaves in > conformance with the spec would be useful, and it would certainly > be less confusing for people who are used to the standard behavior.
I remember we tried to implement this some years ago (IIRC alongside Alexey Klyukin who might remember more details). I couldn't find the thread, but one of the first problems we encountered was that we wanted to avoid storing the text name of the timezone on each datum; we had the idea of creating a catalog to attach an OID to each timezone, but that turned very quickly into a horrid mess and we discarded the idea. (For instance: if a new timezone is added in a new tzdata release, it needs to be added to the catalog, but how do you do that in minor releases?) -- Álvaro Herrera http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training & Services -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
