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

Dave Cridland dave at cridland.net
Wed May 17 22:44:38 UTC 2006


On Wed May 17 20:39:35 2006, Carlo v. Loesch wrote:
> compression is already being used, and it doesn't save you.
> again, see last month's discussion.

I'll take you to task on this one. I'd personally be quite astonished 
if the existing multiple presence stanzas, being presumably 
compressed into a single deflate block, didn't compress very well 
indeed.

To give you an example, compression makes your "expensive" example 
turn into 152 octets. Your "smart unicast" proposal runs to 94 
octets, compressing to 62. To be fair, I think I included the 
newlines from the examples, so the figures aren't exact.

This is using RFC1952 compression - raw deflate blocks, as used by 
TLS. XMPP's stream compression muddies the water somewhat, as it uses 
zlib for some bizarre reason, but in-stream results will be the same.

Bear in mind around 40 octets of TCP/IP framing will be added to 
both. I suspect that for many s2s connections, reducing the TCP/IP 
framing by simply buffering for half a second or so will yield just 
as much reduction in bandwidth - but this is only a suspicion, I'd 
need to see how many TCP packets are involved in s2s connections. 
(Not the data, just the TCP/IP headers, if anyone wants to drop me a 
trace in the mail so I can look).

Dave.
-- 
           You see things; and you say "Why?"
   But I dream things that never were; and I say "Why not?"
    - George Bernard Shaw



More information about the Standards mailing list