W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > May to August 1996

Re: ISO-conformant Dates?

From: Ned Freed <Ned.Freed@innosoft.com>
Date: Sun, 25 Aug 1996 16:50:47 -0700 (PDT)
To: lynx@biol.chemie.tu-muenchen.de
Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Message-Id: <01I8P8AUA38K8Y59WJ@INNOSOFT.COM>
> Quick suggestion:

> There's some ISO standard, don't know which, which suggests to
> use this notation for dates:

> 	YEAR-MM-DD HH:MM:SS TZ

> or something like that.

> This format is very handy as you can use lexical comparison
> to compare times.

First of all, your assertion that lexical ordering suffices to compare values
of this sort is trivially shown to be false. Suppose I have the two values:

   1990-04-04 07:00:00 -7000
   1990-04-04 00:00:00 +0000

A lexical comparison will find these values to be different when in fact they
are the same. If you want to compare time values you have to either normalize
them with respect to zone or else figure out a way to rank times in different 
zones. This takes the problem out of the realm of simply lexical comparison
from the outset, for the format you propose at least.

Second, ISO standards define a zillion different formats for time values, but
as far as I know the one you describe isn't one of them. (It is, however, quite
possible that such a format exists and I am unaware of it, as a complete set of
ISO specifications will easily fill a large bookcase completely.)

The ASN.1 specifications, for example, define universal time, which can be
either YYMMDDHHMMZ, YYMMDDHHMMSSZ, YYMMDDHHMM+ZZZZ, or YYMMDDHHMMSS+ZZZZ.
(There are also some broken implementations around that generate +ZZ instead of
+ZZZZ. Also note the use of two digit years here, and the inability to tell the
difference between a four digit year and the presence/absence of seconds. ASN.1
universal times are a monumentally poor way to represent time values in my
opinion.)

There is also ASN.1 generalized time, which has a whole bunch of different
forms, including YYYYMMDDHHMMSS.SSSZ and YYYYMMDDHHMMSS.SSS+ZZZZ and a bunch of
others. This is a much better way to represent time values, but it is
unfortunately few if any applications use it.

However, I suspect that the format you are thinking of is the one in ISO 8601.
The forms it allows are YYYYMMDDTHHMMSSZ and YYYYMMDDTHHMMSS (the latter is in
local time. This format is used by the VCalendar specification (currently being
considered by IETF Calendaring WG), for example. I think having either local
values or GMT values is kinda strange, but there you go.

In any case, by restricting yourself to either GMT times or else to local time
you actually do end up with strings you can compare lexically. Needless to say,
use of local time is inherently ambiguous and hence not acceptable, but using
GMT always would work and get you what you want.

However, I question the need for any of this. While it is true that the formats
we tend to use for date/time values cannot be compared directly, I remain to be
convinced that this gives enough of a performance gain to warrant changing
things.

				Ned
Received on Sunday, 25 August 1996 17:19:00 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:32:08 EDT