My understanding for poor Jingle support is that, whereas text is easy and low-bandwidth (so you can afford to send everything through a 3rd party server with easy connection semantics), voice and video are much, much much more complicated: not only do you have multiple voice and video codecs with iffy support and stability (sure, even if support library claims to support CodecFoo, it may still fail when passed data from your peer that uses this other library), but the bandwidth requirements means that you now have to deal with trying to directly connect these two clients despite any silent NAT or firewall that may be hidden around. Jingle provided its NAT-traversal algorithm, but even that required network support in the form of "superservers" for when all else failed.