On Jabber/XMPP

I’ve posted a bit a while back about Jabber and my trials with finding a good client, and how I think that jabber is the key to making microblogging really work. As I’ve been dealing with these issues/ideas in greater detail, I’ve had some more thoughts on the subject that I think I’ll collect here.

First off, a subtitle and introduction to XMPP:

The easiest way to understand XMPP (or jabber, which I’ll use semi-interchangeably) is that it’s like email: everyone can run their own server, there’s a specified message metadata (like a header) that makes it possible for lots of servers to talk to each other. Non-jabber IM systems have a central server, which everyone authenticates with and that receives and distributes all messages/statuses between users. So why is distributed good? It solves a lot of scale issues, if most individual servers only need to handle the traffic for a very limited number of users, scaling becomes a horizontal (more servers) rather than a vertical problem (bigger servers.) Also distributed in this case means more individual control over computer usage without sacrificing potentially huge (or at least dunbar’s number sized) social networks.

While this is all very interesting truth be told, this stuff has been around for a decade or more, and for a long time jabber has been the kind of thing that you used a little and then went back to AIM because the clients all sucked, no one really understood or used it, and frankly AIM worked better. In the last year or two things have really started to come together, I think, and XMPP is starting to seem really cool, and really powerful. Truth be told, I think this has a lot to do with critical mass, and the fact that google’s IM platform (Gtalk) is XMPP/Jabber based and interoperable, and so there are a lot more people who are now jabber capable.

I think it also has to do with the fact that jabber gives web based software developers a way to offer push based access without resorting to polling APIs which doesn’t scale well and gets ugly pretty fast.

The breakthrough moment for me--or moments--have to do with multi-user chats. Now the internet has been doing group chats longer, I think, that we’ve been doing one-on-one chats, but IRC--the typical mode--is also centralized, or it’s distributed, in the sense that there are lots of servers, but the servers aren’t interoperable, and it doesn’t scale well at all, and it requires IRC specific software that most people tend to just use (anymore), and that means multi-person conversations aren’t incredibly common-place.

While XMPP multi-user-chats or “MUCs” aren’t too common yet, and (probably) have a host of problems which I’m totally incapable of addressing, I think there are a number of reasons that MUCs are really cool and already beat out IRC in a number of ways. The two biggest “pros” are the fact that users only have to sign into one thing, with an account that most folks already have,¹ and that since servers are interoperable, you don’t have to manage a bunch of different connections with centralized servers.² I’m convinced that the scaling problems of both IRC and XMPP (as technologies,) are nothing in compared to the scaling problems of human-text-based conversations.

And as if that wasn’t enough friend of the blog is putting together the web-based front end/xmpp group chat service that might just make this all make sense to everyday users. While I might not be a big fan of the web-based service for day-to-day computing, there are times when it makes a lot of sense (particularly for capturing new users).

There’ll be more thoughts, as there always are, but I think that does it for now. I try and sprinkle the really-geeky posts in with the more accessible posts because I think it helps me to understand things more clearly, and because I know the geek-quotient varies a good deal. If any of you want more details or have a particular interest in pushing me in one direction or another, I’d love to hear it.

Onward and Upward!