Go to the first, previous, next, last section, table of contents.

Experimental Compilers

Programming languages have been largely Lost In Time for twenty or thirty years: What we're using today is not much better than what Simula 67 was providing a third of a century ago.

How about some fresh ideas?

Muq potentially provides a great platform for prototyping new programming language ideas: The existing virtual machine and the existing support libraries can be re-used rather than re-invented, chopping years off the development time, and the existing base of Muq users and applications provide a testbed in which experimental ideas can be tried out on useful real-life tasks.

How about allowing source code to be in modern HTML instead of archaic ASCII? Then we could write true subscripts and superscripts instead of using ugly circumlutions. We could even include explanatory diagrams in program comments!

How about experimenting in Muq with aspect-based programming? (See http://www.umcs.maine.edu/~ftp/wisr/wisr8/papers/kiczales/kiczales.html.)

I've been arguing for decades now that program source code should not be thought of -- or manipulated -- as ascii text, but rather as sophisticated datastructures something like the program dependency and dataflow graphs used internally by optimizing compilers. Source display should be more like database queries and views than static text, and editing operations should be more at the semantic level of "rename this global variable" than "change this word".

Muq, is among other things a highly programmable database optimized toward large, complex, irregular datastructures with considerable compiler support, so it should make a great platform for experimenting with such ideas!

Go to the first, previous, next, last section, table of contents.