Dear JVP

>>If you build middle-tier components that encapsulate RV's, ultimately, if you need a web interface, you are going to need to convert the data to another format in order to get it to the UI. In this scenario, the RV is an unecessary and a non-value added layer.<<

I *think* you are saying that you have to acquire, affect and share data in an XML/WIDL format or else you have "non-value added layers".

I realise it is the current fashion to use one format to acquire, affect and share data. That's fine in a book or glossy brochure where every customer has the latest database version, 5Gb Ram and no legacy products, but in my world, customers have established repositories full of data in non-XML SQL databases, custom formats like HL7 or even word documents.

In 2002, IME it is mostly quicker to acquire data as a SQL cursor, work on it as required, then convert to XML, than it is to do the whole thing in XML. I note that Rick Strahl has observed the same.

Have you actually *tried* munging big XML datasets? Those things are loaded into memory and they're slow. Cursors are not loaded in their entirety into memory, they are an order of magnitude quicker and indexing the cursor can provide more magnitudes of improvement.

We need to make sure that we don't start adopting "the operation was a success but the patient died" thinking. If we agree to deliver the same result, Customers want cost-effective performance and reliability most of all.

>>Also, if you want to use OLE-DB or SP's, RV's don't scale there either...<<

Yes, and you can't use diesel in a petrol car. Your point is?

>>They don't, nor does anything scale infinitely.<<

You have stated that RVs scaling is finite compared to SP or SPT. In the context of comparison and for the example I gave, my meaning is obvious.