[Standards-JIG] proto-JEP: Smart Presence Distribution

Carlo v. Loesch CvL at mail.symlynX.com
Wed May 17 18:03:36 UTC 2006


Michal vorner Vaner typeth:
| So instead, you want every server or client to take care of creating and
| keeping these multicast beasts in sync. This asumes that each side
| knows, how the multicast list looks like, each one keeping a cache. You
| need the traffic to keep them in sync instead of the presences and make
| id difficult.

Wow amazing.. amazing visions.. where did you see all this complex
protocol keeping databases in sync on all servers in the little JEP
we wrote? I mean, hey, this reminds me of IRC. IRC indeed keeps all
information about everybody on every server. But no, our JEP only uses
the information which is already there. Each server already KNOWS who
of his servers are subscribed from and to somebody else.

There is a misunderstanding in your thoughts making you think that if
A is subscribed to B on server C, then servers D and E would be put in
the know about it. Why should that be?

| And another thing, what if capulets does not have an antivirus, their
| server gets attacked by a virus, so they just erase it all and start
| again. Montaque server just nicely things they have the list and send
| presences and poor Juliet does not know Romeo is waiting for her to
| write.

Brilliant.. This paragraph is really a fascinating continuation on the
wrong train of thought. But let's make something useful out of it.

Since IRC is a bit like you say, if an antivirus attacked one IRC node,
and the IRC node had a steady reliable database of people like Jabber or
PSYC, all of IRC would have a serious problem. That's one out of many
reasons why IRC has no steady database. Thus, it always has to keep things
in sync with every body else, which amounts to its greatest scalability
problem. I remember when we made IRC statistics we found out IRC is by a
far higher percentage busy keeping all data in sync than actually pushing
around conversation.

Why speak about IRC? Well this post just brought me to it, but it is
good to mention that learning from the good and bad of IRC can be very
helpful in designing a new chat technology.




More information about the Standards mailing list