From: Tom Lane <tgl@sss.pgh.pa.us>
To: "David E. Wheeler" <david@justatheory.com>
cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: Inconsistent Parsing of Offsets with Seconds
In-reply-to: <A1454D02-F747-4A1E-8129-F7E940C7E4EB@justatheory.com>
References: <A1454D02-F747-4A1E-8129-F7E940C7E4EB@justatheory.com>
Comments: In-reply-to "David E. Wheeler" <david@justatheory.com>
	message dated "Sat, 22 Jun 2024 12:25:29 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-ID: <788147.1719076559.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Sat, 22 Jun 2024 13:15:59 -0400
Message-ID: <788148.1719076559@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/788148.1719076559%40sss.pgh.pa.us>
Precedence: bulk

"David E. Wheeler" <david@justatheory.com> writes:
> The treatment of timestamptz (and timetz) values with offsets that inclu=
de seconds seems a bit inconsistent.

It's hard to get excited about this.  Per the IANA TZ data,
nowhere in the world has used fractional-minute UT offsets
since 1972:

# In 1972 Liberia was the last country to switch from a UT offset
# that was not a multiple of 15 or 20 minutes.

and they were twenty years later than the next-to-last place (although
IANA will steadfastly deny reliability for their TZ data before 1970).
So timestamps like this simply don't exist in the wild.

> The corresponding jsonpath methods don=E2=80=99t like offsets with secon=
ds *at all*:

Perhaps that should be fixed, but it's pretty low-priority IMO.
I doubt there is any standard saying that JSON timestamps need
to be able to include that.

> I see from the source[1] that offsets between plus or minus 15:59:59
> are allowed; should the `OF` and `TZ formats be able to parse them?

I'd vote no.  to_date/to_char already have enough trouble with format
strings being squishier than one might expect.

			regards, tom lane