Dynamic languages in ASP.NET performance

In The New Dynamic Language Extensibility Model for ASP.NET, David Ebbo claims statically compiled languages often loose their natural performance advantage during the lengthy and costly process of the HTTP request.

This is, to some extent, true. Let me present you with some arguments about when this claim may fail.

First, let's not forget MTS and Microsoft MTS evangelization. Back in the late 90s, Microsoft convinced many of us (myself included!) to use MTS for data tier in general, whether multiple database access was involved or not. They also claimed that the out-of-proc and MTS overhead was limited by the rest of the pipeline, and they were wrong.

Now for simple demos, I believe David has a point. But for simple demos, even Microsoft demos work, and we wouldn't want to build most of our corporate over those, would we?

My worries goes to the layers we often need to produce when architecting corporate applications. We have to build instrumentation and exception management, we have to enrich the ASP.NET controller and we want to know the soonest possible (even before unit and integration tests) if something breaks.

For last, let's not forget next version of Perl, the duct tape of the internet, the king of dynamic languages, who's strategy passes along with bytecode compiling and the addition of static types.