Well, I think it is horses for courses. And I get the impressions Steve does intend that as well. Some of the new things in SQL 2005, however, may mean you don't want to wait (.. or didn't wait, as in our case - we've been on SQL 2005 since... 2005). That said, we also have several SQL 2000 servers, and some 7's as well. Upgrade where you need to, for good reasons, is my opinion.

Some of the performance and stability issues we (by we, I mean our core client) had was around a massive caching system that aggregates and presents data from several hundred SQL servers, and then makes this available to our customers (somewhere in the order of 500 million to 750 million 'tuples'). Towards the 2nd half of 2006, the SQL 2000 DB cluster had frequent problems - to the level that the cluster was failed over on a near-daily basis. Now I am sure many people will think that this is due to bad code, I should point out that what was done here was reviewed by MS, and was considered by Ms as a very good, innovative solution to a very challenging problem. After upgrading from SQL2000 to SQL2005 (with virtually no code changes), this server now runs at an average CPU of around 10% and has not needed a fail-over in 9 months.

Not everyone will have these needs, but the performance improvement can be significant, as can the 24x7 enhancements.

I should also say that SQL 2005 is not the panacaea for all SQL problem, either. In some areas, I'm not sure it is that stable a product - SQL 2005 SP2 and SP2a debacle a prime candidate. We still have assertion errors on 1 of our (5) live SQL 2005 server clusters, and have had them for 2 weeks now... we are on build 3175 with no resolution. Fortunately, is doesn't appear to have an impact (yet) on our service, but it is very concernign to see 80-100 assertion errors a day, on a server.