[JDEV] Message timestamping and route tracking

[Benjamin Holzman]

Daniel Arbuckle wrote:
> 
> Make tz optional. If it's not there, no problem. Also, tz should be
> formatted as a number of hourse to add or subtract from GMT rather than by
> name or something.

Er, if it can *sometimes* be there, then every application will have to
deal with it.  Consider the common case where you want to sort a bunch
of messages.  If some of them might have a timezone, you have to convert
each timestamp to a neutral format; say, GMT.  Why not just have the
timestamp be in GMT to begin with?  If you want to see timestamps in
your local time, that's the job of your client to display them that way.

> 
> What this adds is the ability to know the local time when the message
> passed through each server. For relaying servers, this isn't valuable
> information; for the sender's and receiver's home servers, it is.
> 
> What would be the point of adding geographic information? This is all
> about time, and an offset from GMT is the simplest universal way of
> declaring a time.

I respectfully disagree. I suggest that *GMT* (or some other suitable
frame of reference) is the simplest universal way of declaring a time. A
timezone implies longitude to a certain degree (China notwithstanding),
which is why I thought you wanted to capture geographic information.

-- 
|        Ben Holzman                       bah at orientation.com       |
|      orientation.com                     Tel: +1 212 966 5553 x219 |
|  Senior Software Engineer                Fax: +1 212 966 5554      | 
$ perl -l040e 'print ucfirst for reverse qw/hacker perl another just/'