From: Russ Allbery (rra@stanford.edu)
Date: Wed Apr 02 2003 - 00:18:27 CST
Tsuyoshi Takada <acroyear@gmx.ch> writes:
> Greetings all,
> I think the legacy Date format like the following is not systematic.
> Date: Fri, 1 Nov 2002 15:21:25 +0000
> This format is complicated especilally in text editing or statistics,
> because the month is not numerical but strings, and the conceptual order
> contains both "from small to large" (day - month - year) and "from large
> to small"(hour - minute - second).
I'm sympathetic to this, very much so. I really hate the e-mail date
format. However, this is one of those things that's just not sufficiently
broken for anyone to go through the extreme pain that would be required to
fix it until the day in the indefinite future where we junk the current
message format entirely and replace it with something completely new.
Short of switching to a new format, it's just too painful to change
something like this without sufficient benefit. Adding a new header won't
get you there, unfortunately, because you'll discover that you'll never be
able to actually stop generating the old one, and you'll have to generate
both headers forever, which is more of a pain than we have now.
The best way to think about this, in my opinion, is to view the format of
the Date header as an unambiguous, if rather odd, machine encoding of a
date, and just write the parser to parse it once (or borrow someone
else's), convert the date to a better format internally as soon as you see
it, and then not worry about what it looks like in the message.
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>