Re: Some clarification about TIMESTAMP

> There are any number of
> server-side settings that can affect the interpretation (and display)
> of your data.  Datestyle for example already renders this position
> untenable.

What makes me a little uncomfortable in this assertion -and in many
parts of PG docs-
is that emphasis put on what "is displayed", as assuming that I will
be using postgresql
in the server, and using the command-line psql. But of course one
frequently (mostly?)
access the DB remotely and from a client interface (eg. JDBC), one
would say that the
display/interpret (from to a string) ocurrs normally in an upper
layer, not in the DB.
(I suspect, from some peeking at the source I did once, that
internally the "canonical"
representation of values in memory is as a string, the same that it's
displayed/parsed
in the psql - and so, even if I don't use the psql CLI, the convertion
to/from string happens
the same internally - is this true? that's far from obvious for me.
Say: when I query a timestamp from the DB via JDBC to display in a jsp page, it
first converted from the binary store format to a psql-like string,
then the jdbc driver
parses that string to convert it to a Date java object, and the
finally Java converts
it to a string again? Is that so?)
> It did, but extract(epoch) assumes you want a distance from the real
> Unix epoch, so it takes the timestamp as being in local zone.  AFAIR
> there isn't a function that does exactly what you seem to be thinking
> of.

Weel, it seems that if I want that timezone-agnostic behaviour,
so that extract(epoch) always returns the same integer for a given stored
value (and different server-configred timezones) I must use (no very intuitive)
a TIMESTAMP WITH TIMEZONE.

Regards

--
Hernán J. González
http://hjg.com.ar/