Although I complain that IM stands for "interruptive messaging", I do like to chat with people because I recognize the value of continuous partial attention (often people need to ping me in order to keep going on a task they're working on) and because, more generally, talking with people is in large part how I work in the collaborative community we call Jabber.
Case in point: earlier today I had a longish chat with Tom Germeau, who is working on Mozchat. Tom basically asked me what features my dream Jabber client would include. Now, I don't lie awake at night dreaming about Jabber clients, so his question brought me up a bit short. After we talked a bit, I came to the conclusion that I'm not interested so much in the one kickass Jabber client to rule the world (though a really nice Mozilla-based client would be quite cool), but in growing the Jabber ecosystem by integrating Jabber more deeply into the rest of the Net. By which I mean things like:
IMHO, we see all these social networking things going on but in my experience they're not very social, not dynamic, not realtime. Take Orkut. My experience was that it was fine a fine idea, except I had to go there to do things, and I don't go places, I need reminders that something interesting is happening over there because my attention is so easily pulled away. And IM is a great way to push things out to people and grab their attention.
Personally I'd also like to see bits of Jabber get built into more apps, and one way to do that would be some kind of Jabber daemon that runs on your machine so that other apps can tie in with that and make use of the connection for realtime communication and presence. One simple example might be little presence icons in Thunderbird (or other mail clients) based on the Jabber-ID email header. Other examples might include Ebay auction notifications and the like.
For what it's worth, my hope is that XMPP will become an open standard for passing around any kind of (smallish) XML formats, presence, little workflow apps and voting bots via x:data, pubsub stuff for various payload types such as geolocation and tunes, etc. Granted Jabber technologies already do all this, but they're not as widely deployed as they could be (though that is changing every day as more and more people build and deploy Jabber-based applications). The key is that XMPP servers are pretty much payload-agnostic, so we can let the servers just route the data around, which enables people to build fun and interesting services.
Does this mean we're trying to take over the world? Although we used to joke about world domination, I would say no. My understanding of the overall Jabber goal as originally set by Jer, which I guess I inherited from him, is not that we're trying to get rid of email or HTTP or SIP or anything else, because those technologies have their own strengths. Instead, our goal is to offer a fully open technology for real-time communications and let developers innovate around that. We're not trying to do it all because we can't. We're just providing an open, extensible alternative that will enable people to innovate at the edges without asking anyone's permission, extend it how they want, build things we haven't imagined yet, and (we hope) do so in an open manner that enables freedom of conversation and seamless federation.