Walter,

Once you hit a 1-M scenario, for ease of development the front end has an obvious advantage- all the rows are right there and you can iterate your inserts in a loop or via an append. Passing to a middle tier or SP requires some sort of 1-M transport mechanism which in 2006 is XML. XML is described as many things but IMHO it is a partial industry response to the need to safeguard important data in an increasingly connected world. You'll agree that if you are responsible for a shared pool of data on which people rely, you'd want to have not just responsibility but also complete authority over it. So you can't allow people to insert data as they please and hope they're doing it responsibly. You could use middle tiers or encapsulated data classes to enforce discipline, but SP means a centrally located manager or technocrat can exert absolute centralized control without the risks of wrong versions of classes, hacking or other problem activity outside their remit.

There is another issue: secrecy. If you have devised a remarkable new algorithm or process that gives you a business advanatage, competitors are keen to mimic your success. The easiest way to do this is to get hold of source code or personnel who have worked on it. SP or centrally managed business tiers mean the app is never in the competitor's hands for decompilation, and it is easier to control access to the source as well.

Despite claims to the contrary, there is nevertheless *plenty* of thinking out there that if munging is needed by a particular PC, why wouldn't you perform the munging on that PC, especially if it is a typical modern workstation that is running at 2% capacity most of the time? So we start to see initiatives like Linq and mini-databases that reside on individual PCs. And if you control both the database and app end for a system that is being used by a customer for their own data, then you have a lot more flexibility.

You know, IT seems to move in long, dignified cycles. My first PC was a TRS-80 whose operating system and Basic interpreter/compiler was on a 16K chip. I was most impressed when the PC's OS moved to a spinning disk so that I could easily implement newer versions. Then the OS got bigger, larger and more incomprehensible when things went wrong. Now in 2006 I am delighted at the thought of a server whose Linux OS is entirely on a CompactFlash, giving me huge reliability advantages over the spinning disk version. Which gives me the opportunity for a terrible pun: "what goes around comes around".

;-)