[jlg23]

> Looking at modern day web development we are seeing the opposite, where the program is many times bigger than it's output, suggesting to me that the modern frameworks do not come from any sound theory.

I tend to agree, most "modern [web] frameworks" were created as hacks to ease some other development pain and only later got polished by the application of some sound theory (or simply by enough development/use-cycles).

OTOH, since frameworks must be generic enough to, well, act as a "frame" for the actual application, one could also argue that more can be gained by some JIT-mechanism that optimizes according to the current usage pattern. (Somebody, IIRC at MIT, published something along those lines just a few weeks ago.)