Triangulating position from wi-fi signal strength instead of GPS

Pointed out in an IEEE Newsletter and they saw it on this blog.

They say it's good to 20m in dense urban environments and doesn't have the multipath issues of A-GPS. Since it's going on signal strength, I'm guessing you'd probably want 4 sources for a reliable fix (two signals would give you two equally probable locations, a third would say which of the two, a fourth would make me happier-- especially if the other signals are coming from the same general direction). This does limit you to dense downtown areas -- but that's what it's made for. The blogger who first pointed this out mentions people moving and taking their transmitters with them.
Update: John Krumm from Microsoft Research recently presented a paper on this at MobiSys 2005, the Third International Conference on Mobile Systems, Applications and Services, held June 6-8 in Seattle.... apparently signal strength doesn't work as well as "other aspects of signal quality" hmm? The paper will appear in ACM here for subscribers.
Update 6/30: I just noticed that the July 2005 issue of IEEE Signal Processing (link for subscribers to Xplore) has a large special section on positioning in wireless networks.