Also,
you need to do that first when you actually have external customers accessing your systems, so you are fine without registration for testing, development, beta test etc. At that point, you should have a pretty good idea on if it is worth taking 5 min to get a closed-source license. And the free server has no limitations except scaling out whatsoever ... pretty good deal IMHO.
Hi there,
being the author of this article, let me just point out that I am _not_ an "social media consultant" but part of the scientific research carried out at www.linkedprocess.org. This article is merely trying to summarize and relate different computing approaches like Chat, GWave and others. Much of what is pointed out in this article is already either in the XEP (which is, others are actually extending the protocol and requesting these extensions to be standardized) or projects are exploring these scenarios (see the links in the comments and the article itself).
So, I don't think this is reality distortion, Google Wave and others placing their "next generation conversation" bets into this basket is IMHO a very fundamental endorsement.
XMPP is not the tightest of protocols, but it adds a lot of value in the outlined scenarios since you get internet-scale discovery, status handling, routing infrastructure and handling of extensions and custom payloads for free.
The XML overhead is IMHO only evident if you are either very low on bandwith (like mobile networks before flat rate and 3G) or parsing the XML yourself (not using the excellent client libs out there), so I think that is acceptable. I agree with some of the other comments that there never will be an ideal protocol, but this actually works, has exisitng deployments, clients and libs, and there are open servers and specifications to roll your own if you like.
/peter