Re: date parsing on "bite-sized tasks" list

Greg Stein wrote:

>On Wed, Aug 22, 2001 at 08:42:22AM +0200, Branko Hibej wrote:
>
>>...
>>My suggestion would be to do the following:
>>
>> * For input:
>> * Always acept ISO 8601 dates, times, and intervals
>> * Accept dates conforming to the current locale settings (as
>> modified by --locale), wherever those can be guessed
>> * Accept unambiguous "common" designations, such as
>> "yesterday", "next monday" (and possibly "easter" and
>> "sammain", if you can work those out. :-)
>>
>
>Agreed.
>
>> * For output:
>> * Write dates in the format defined by the current locale
>> * As a possible exception, with --locale=C, write dates in ISO
>> 8601 format.
>>
>
>For user-visible dates: yes. For internal date strings, we should only use
>ISO 8601. In particular, I've noticed that SVN/entries uses a wacky format:
>
> Tue 21 Aug 2001 20:00:19.000000 (day 233, dst 1, gmt_off -25200)
>

(Having the timestamp in this particular format helped me find and fix
problems in APR's date handling fns by the way.)

>That is just overly complex and will be difficult for "foreign" utilities to
>deal with.
>

What foreign utilities? I thought we agreed that the WC metadata was
private, and any utilities should use libsvn_wc to access it.

>For example, "Tue" and "day 233" are useless since they are
>derived values. And the zone stuff is all useless since stored times should
>be in UTC.
>

Why in UTC, if it's in WC?

> ISO 8601 specifies a well-defined format for this stuff.
>
>See:
>
> http://www.cl.cam.ac.uk/~mgk25/iso-time.html
> http://www.w3.org/TR/NOTE-datetime
>

But yes, in general I agree with you. I've said the same several times
-- even posted the first of these two URLs before. :-)

-- 
Brane �ibej   <brane_at_xbc.nu>            http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org