A strong second on that recommendation to hold out till 2008.

My only disagreement with anything Steve wrote is regarding the stability of SQL Server 2005. Yes, the data engine of 2005 is surely stable, but the development integration -- where they seemed to focus -- is less reliable than it should be given the beta cycles they went through. The best example of this is the horrendous management tool interfaces, which on many development boxes were seriously hamstrung. And requiring a development tool upgrade to VS 2005 just to get at some features was a problem in terms of deployment, since VS 2005 was not backward compatible. It produced an added layer of swkward management, and that affecte dproductivity.

I'm not saying SQL Server 2005 was a bad release, though, just that when you analyse their development process, beta cycles, and so forth, it just wasn't really a solid release. The same is true of VS 2005, which wasn't really usable in many scenarios until SP1.

My biggest fear for 2008 is that a trio of upgrades will create  a slew of isolated platform bugs, and an unpleasant batch of integration-bugs. MS needs to consider slowing down and committing to point releases and optimisation for a few years after 2008, or it will cut their bottom line.