I think comparison anything on Windows or x86 to the mainframe world is rather foolish. That is a comparison that holds no water without much further proof.

I can accept that legacy code is important and must run well. In fact, I can even accept that legacy code must increase in speed. I don't think you can make a compelling argument in the x86 world for legacy code being of equal importance to modern code when it comes to execution frequency.

Think about how generous MS is with support for older OSes...how many mission critical NT3 installations exist do you think?

DK