[1597] in Discussion of MIT-community interests
Re: [Mit-talk] Fwd: [Jabber Project Announcement]
daemon@ATHENA.MIT.EDU (Greg Hudson)
Mon Jul 18 17:06:23 2005
Date: Mon, 18 Jul 2005 13:37:46 -0400
From: Greg Hudson <ghudson@mit.edu>
To: Natan Cliffer <natan@mit.edu>
In-reply-to: "[1595] in Discussion of MIT-community interests"
cc: mit-talk@mit.edu
Errors-To: mit-talk-bounces@mit.edu
> With the right amount of fudging, is there any reason why a
> jabber-based system couldn't have a wrapper that looked just like
> zephyr?
Possibly. The Jabber protocol contains the concept of a thread
identifier, which could be mapped onto the Zephyr concept of instances
for the usage model of classes like sipb, where users generally
subscribe to the entire class. I don't believe existing clients
display the thread identifier, though.
The usage model of class message, where users generally subscribe to
specific instances but are allowed to subscribe to all of them, would
require some extensions to the chatroom protocol.
Brian Pepper wrote:
> The END of ZEPHYR?!?!?!?! </chicken-little>
For the moment, Jabber is being positioned as a competitor to Zephyr,
not an immediate replacement. However, there are a bunch of ways
we're not happy with Zephyr:
* Reliant on Kerberos 4
* Poor-quality code base
* Code base maintained only by MIT and CMU
* Usage mostly limited to Unix users
* Not NAT-friendly
* Clients not terribly user-friendly
* Terminology ("classes" and "instances") not very intuitive
* Poor support for communicating outside of MIT
* Subscriptions can be lost without notice to the client
* zwgc's pop-up windows not supported well by metacity
So while we have no specific plans at the moment, it seems possible
that at some point well down the road, we will look into phasing out
Zephyr. There are also some things we like about Zephyr which we'll
miss, some of which we may get back through development effort:
* Very high availability (clients not disrupted by downed server)
* Good client support for being subscribed to many classes
* Client display of instances when used as thread identifiers
* Terminal clients
Mapping Zephyr onto Jabber as Natan suggests is an interesting idea
for building critical mass for Jabber, and also might help to smooth a
transition away from Zephyr.
_______________________________________________
MIT-talk mailing list
MIT-talk@mit.edu
http://mailman.mit.edu/mailman/listinfo/mit-talk
