<video> isn't ready yet because the implementations are very young, and they still have pretty basic bugs. We don't yet have a good comprehensive test suite for the API, and the API in fact is still in active flux (I'm literally changing parts of the API as we speak -- I have an incomplete edit that changes how the timed track API works that I'm still working on). And of course there's the issue of the codec, which is still unresolved.

Deploying a site with <video> and Flash is twice as much testing, implementation, and encoding cost as just <video> or just Flash.

Using <video> today (late 2010) is fine for experimental purposes, demos, or for people who are on the bleeding edge. But for mainstream sites who need to reach even people who use IE6, or who don't have much in the way of resources to spend on this kind of stuff, or who simply aren't interested in being the pioneers of new Web technology, then I wouldn't recommend using it yet.

This kind of stuff moves really quite fast, though. Six, twelve months from now? Who knows. Maybe it'll be deployed widely enough and with good enough quality of implementations that using a proprietary plug-in for video will seem positively anachronistic.