Dear JVP

>>I understand that you are constrained by legacy issues, and that is an important issue to consider.<<

Yes, customers do have valuable source data in legacy systems, one has to be sensible.

>>If there is data to munge, I would look at what I was doing. If it was aggregate reporting, I would consider Analysis Services as a way of achieving the same goal. Or, I would consider server-side components or SP's. IAC, I would not consider downloading huge chunks of data to the client...<<

Are you sure we're talking about the same thing when we say "Web Services"?

The Web Services world is quite different from traditional development where one vendor or party controls the whole show and can distribute processing between server and client.

The "client" of a web service can use any product they like to make their request, for whatever reason they may have.

I cannot possibly instruct the consumer how much data they need or tell them to process stuff on my server. I do not know why they want the data, the only "rationing" mechanism I have is cost.

>>So you recommend downloading large #'s of records to the client and process on the client?? Orders of magnitude better? Maybe better than XML, but it is not the optimal way of doing things...<<

While you might declare rules about what people can request, there may be *excellent* reasons for a consumer to request data more than you might feel is efficient. Example: a hospital in the UAE requests a VAST data collection... you do not know it, but they have a collapsed holidaying dialysis patient and need to load the patient's full record into their Web EPR ASAP to expedite multiple specialist opinions.

Will your system refuse because it decides they don't actually need all that data? Or will you insist they guess date ranges and make multiple requests to avoid overloading your middleware? Or will you tell them they should have used Analysis Services? (!) Or will you abandon the middleware idea and load the server instead?

All because you have chosen a technology that does not scale?

No. you can set limits on data but it has to be a limit that makes sense wrt the business need. So you could say "only one patient request at a time". Limiting by volume would not make much sense in health.

Regards

JR