| Dave Cridland wrote: | > I basically refuse to consider any solution seriously unless you're | > measuring the effects of it post-compression
Philipp Hancke typeth: | you are assuming that we are talking about a mere optimization of | interdomain traffic which can be measured in byte/second? I presume the amount of xmpp parsing - not just the amount of data - brings in some unnecessary load on servers. Let's optimistically consider the case s2s federation has to deal with roughly half as many stanzas as before.. wouldn't that be a great relief, no matter how much compression and xml parsing tricks are applied? Maybe a more thorough effort in counting the number of redundant stanzas would be useful to find out how much applying repeaters (or whatever else) brings. I know real multicast is even further down the road for xmpp, but that would relieve the load on "source" servers enormously if they no longer need to get in touch with each and every subscriber's server. So, Dave, if you think the repeaters strategy isn't good enough, what about going for the real deal and come up with a multicast plan?
