Hello Steve,
I agree with you that while excitment sells it doesn't necessary mean progress. And not everyone wants to blow the same whistle and reach for the newest brass ring. But when it comes to things like matrix work benches there seems a need for a reality check too. What you are encouraging is to go where there is absolutely no intent. Yeah, you can play around. You can also try to manage an enterprises data with fortran. So it's not surprising that some spacecadets think nothing of using the server for factor analysis. If this is for 'real' and for 'real' data, then here's the check mate - that's crazy. Professionals know the tools for their trade, or should. You don't play chess with checkers. If you want to deal with matrices or multivariate analysis you go with the software where that's the intent. You check out the IMSL library, SAS, SPSS or BMDP. They've been around a hell of a lot longer than sql server and their stuff is burned in. You don't go cobbling together select and backdoor update statements. MS users should learn more about broads, as in horizons.
best,
steve dassin
P.S. If you want real arrays in real data management you need look no further than the combination of Dataphor/Sql Server.