[Standards-JIG] Client Capabilities (rant)
trejkaz at trypticon.org
Sun Nov 20 23:09:40 UTC 2005
> If I only have 1 client in my roster, and that person has 10 ext items,
> it will take me 11 requests go get all their features. One request for
> the base node (client#version), and 10 additional requests (one for each
> ext item).
Or just one request to the user in question.
> It's not until we have a roster with multiple users all running the same
> version of the same client that we start to see a lowering of disco info
Well, that much is certainly true. And I've been wondering about this
caps stuff for quite some time due to this phenomenon. Sure, a lot of the
users on my roster use Psi because that's the circle I seemed to end up
in, but if I casually browse around, I see that some people are on various
test versions, some are on the previous version, one guy is still on
0.9.2, and then there are a few people on Adium, all of whom are on
different versions of that as well...
Not only does it not decrease the number of disco#info requests until you
find two of the same client, but until then, it actually costs you more,
because you're receiving this extra element in the presence.
> Now, if we cache values across sessions that we see a dramatic falloff
> in disco requests, but the JEP doesn't mention that, and I don't know of
> any clients that do it (although I haven't looked for that either).
That would seem to be a smart optimisation anyway, so perhaps the JEP
should mention that if it isn't there already. It seems like a reasonable
thing to do... but of course, clients already cache things like the vCard
without being told to by the JEP.
More information about the Standards