I completely agree with Brian.

There are a number of issues here...

1) Comparisom with other vendor's release stategy.

My main experience is with DB2.  When IBM are near to making a new DB2 release RTM (GA in IBM terminology), they will make known to technical conferences the major line items in the subsequent release  (Vnext), and those line items that will not be considered until after Vnext is RTM.  This gives about a 6-year plan for the product's future.  They also have about a 2.5 - 3.5 year release cycle and regular (approx quarterly) service packs.  The shorter release cycle means an easier upgrade path, with the +1 release supporting everything that has not been deprecated in the current release.

All this allows sites to prepare far more for what is coming, and the regular SPs help greatly in budgeting manpower for maintenance.

2) Credibility of vendor

When a vendor makes it known they hope to release a major update in 3 years, and it actually takes 5 (e.g. SQL Server 2005), you start to wonder about their ability to deliver for the future.

If Microsoft continue to deliver major upgrades a year or two after making their initial plans known, it will have a corrosive effect on their market share, regardless of final product quality.

3) Size of feature set

SQL2005 is massive compared to SQL2000, but SQL2000 is not the main competition.  Compared to its main rivals, SQL2005 will deliver VERY little that will not already be in other products.

For SQL2000 people, the feature set size is a problem, especially for DTS.  I think it is a major error on Microsoft's part the SQL2005 DTS does not have source compatibility with SQL2000 DTS.  Run time compatibility only is simply not good enough for adjacent releases.  We will have to re-engineer all our DTS at the time of migrating to SQL2005, to ensure that any further troubleshooting or maintenance of a single DTS package will be compatible wih the other packages it works with.  This will add perhaps UKP 200,000 to our SQL2005 upgrade costs, due to the amount of DTS we have.

4) The SQL2005 legacy

The frightening aspect of a 5 year wait for SQL2005 is the prospect of a further 5 year gap for the next release. 

In a number of areas, SQL2005 technology will be inferior(*), and while its rivals move further ahead with their 3-year release cycles we will be stuck with what we have until maybe 2010.  By then, we will either have a release that makes SQL2005 seem minor, or SQL Server will remain permanently and visibly many years behind the times.

(*) Areas where SQL2005 is inferior include continued index locking for insert/delete activity, single dimension only index clustering, slow adoption of SQL standards changes, etc.