[JDEV] Message timestamping != User info

Lazarus Long lazarus at overdue.ompages.com
Tue Oct 5 17:47:00 CDT 1999


On Tue, Oct 05, 1999 at 08:47:15AM -0700, Jon A. Cruz wrote:
> >         Forgive my foolishness, but couldn't we use something like they
> > use on IRC for ping times?  They use the number of seconds from Jan 1,
> > 1970 (or something, I am not sure whenthe date is) GMT to calculate dates

> Well, one problem is that number of seconds will wrap at the year 2035 or
> 2038 depending on your platform and use. But more than that, whether we send

Use 64 bits rather than 32.  Two octets is surely more efficient than
the "human readable" formats others have proposed.  Isn't the intent to
have clients perform translations to "user friendly" formats if that is
desired?  Why should servers (and pipelines) be saddled with transporting
more data than necessary with every message?  User interaction is
intended for the client level, not the server level, and "human readable"
formats are better provided there, in whatever format the *individual*
user prefers, based on client configuration.  (Example, 01/02/03 means
different things to different users, and translations per-user at the
*client* level will prevent misinterpretation.)  Anyone expecting
that Jabber will be relevant when a 64-bit field is filled needs
some perspective (and being open source, migration to larger fields
would be trivial as the date approaches.)

UTC is called Universal for a reason, regardless of whether humans in
certain countries have caught on or not.  Seconds since the beginning
of the epoch is well-supported in languages with good portability (such
as C.)  A single numeric value is certainly easiest for sorting a dozen
messages by message-generation time, rather than juggling multiple
timezones, et cetera.

A timestamp on a message should not need to indicate geographic
information (such as *user's* timezone) which is more pertinent to the
user, not the message, IMO.  I really don't see a significant use for this
information anyway, but if desired, I feel it would be more appropriate
to provide it elsewhere, not with every message sent.  (Do I really care
where someone that writes to me lives?  If I did, wouldn't I already
know that?  Conversely, do I really desire to divulge my location to
each and every person to whom I send a message?)  Believe me, Jabber is
already being looked at by privacy advocates and there will undoubtably
be anonymizing servers available when the time comes.  Not only is this
data irrelevant, but automatic disclosure of it is undesirable by many.

Summary: *when* I wrote the message, not *where* I wrote it (no matter how
vague) is what is relevant.  Don't forget, many of us don't choose to keep
"daylight wakefulness" schedules either, so local time is irrelevant if
you don't already know the person generating the message you receive,
and if you do, then a timezone field is just wasted bandwidth.  Why not
add a hair-color-at-time-of-transmission field?  (That last is humour,
if not apparent.)

-- 
Please encrypt all mail whenever possible.  The following Public Keys
for Lazarus Long <lazarus at overdue.ompages.com> are available upon request:

  Type    Bits/KeyID    Fingerprint        (GnuPG (GPG) is preferred.)
GPG/ELG: 2048g/DE3E078A 0023 E86F A45A C46D 8243  1A7F 6FF2 84C1 3F5B F7CD
GPG/DSA: 1024D/3F5BF7CD (none for DSA keys)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 244 bytes
Desc: not available
URL: <http://mail.jabber.org/pipermail/jdev/attachments/19991005/cd631797/attachment-0002.pgp>


More information about the JDev mailing list