[Standards-JIG] Trying to understand

Dave Cridland dave at cridland.net
Thu May 18 22:53:55 UTC 2006


On Thu May 18 22:00:26 2006, Matthias Wimmer wrote:
> My point was not this representation, I just used it as an example. 
> My point was: Couldn't we agree on only sending out one stanza, but 
> let the source server include the addresses to which they should be 
> sent.

I suspect that however you choose to represent this, the figures 
would work out roughly the same in the end. You're carrying the same 
information whichever way you represent it.

> If we would care about each bit on the wire, we would not use XML 
> but a binary protocol, and we would have to use shorter tags. But 
> that's not the philosophy of XML.

I'm not really sure what you mean here. I hope we do care about each 
bit on the wire, for two reasons. Firstly bandwidth costs, and 
secondly fewer bits means more speed. If we don't care at all about 
bits on the wire, then really we shouldn't be having this thread at 
all.

Whether you happen to express in XML or not isn't important here - 
obviously XML itself is quite verbose, but in principle, the 
difference between representations ought to be small once compression 
is applied. By introducing more framing to enable a smaller 
uncompressed representation, you actually reverse the effect once the 
compression is applied.

And you really do want compression, all the time, everywhere - it's 
an instant saving of time and money, and saves a considerable amount 
of energy trying to define the perfect XML representation, as you've 
probably seen by now.

For stanza level savings, you'd really be better off examining why 
there is such high amounts of dissimilar presence traffic, and see 
what can be done to decrease that.

I suspect looking at ways to make things like JEP0115 and other 
presence stanza payload data more efficient might help, as well as 
experiments with presence damping.

The Smart Presence Distribution protoJEP is also looking in the right 
direction for savings, the problem there is the associated security 
concerns. I prefer the concept of defining and manipulating an 
address list on the remote server, which ought to yield similar 
savings, and lessens the security concerns. It also opens the door 
for more interesting use-cases, such as sending extended presence 
information only to those contacts who actually have the client 
support required to see the information.

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