>Again, this is anecdotal evidence which does absolutely nothing to argue the anti-RV sentiment.

PMJI but I've written two apps that used both, and although I can't provide hard numbers, RV's were definitely slower than using SPT/SP. Given time, I could probably work them up, although my current employer might take a dim view of my spending the time on it.

The first app still uses a mix of parameterized RV's (mostly for data entry) and SP's (mostly for data presentation/reports). This app is currently fielded in over 70 offices across the US, user count is probably over 200. The users complained from nearly the first day about the speed, which was tied directly to using RV's, and as the user count grew, speed problems grew. It is my understanding that a prototype has been rewritten to use SPT/SP's exclusively and gives significantly better performance.

The second app was a small data conversion device I built to port data from VFP tables into SQL Server. The table being converted was 100K records. Out of curiosity I wrote it three different ways, one to use a remote view, one to use ADO, and one to use SPT (the update string being built on the fly). By far, SPT was the fastest, on the order of three times as fast. Furthermore, it could be started and stopped, whereas the RV-based version tried to bring down the entire result set when trying to do an update (even if opened NODATA).

I will be the first to admit that the RV's in either app may not have been properly constructed; time pressures on getting the first app running were incredible. As for the second app - well, once I was able to show that the SPT version was faster, I stopped fooling with RV's.

Anyway, that's just my experience.