Re: The Free Semantics Foundation

[Stefan Monnier <monnier-CRDzTM1onBSWkKpYnGOUKg@public.gmane.org>]

> I have written this as a sort of manifesto for a project. The idea is to
> develop software for automating programming. If we can automate the
> production of concrete implementations of communications protocols, device
> drivers, language interpreters, etc, then we can change and combine such
> implementations much more easily. We could also secure systems by design:
> if all the code on an exposed interface in a communications or an operating
> system is automatically generated, then we can ensure that buffer overruns
> etc can't happen.

People have been doing that for ages.  E.g. describe your protocol in
some language, then auto-generate an implementation from it.
The "some language" needs to be precise enough that an implementation
can be generated from it, so typically this language ends up looking
pretty much like a programming language.  Maybe specialized to the task
at hand (that's called a Domain Specific Language, and is routine in the
Lisp world) or with a slightly different feel (e.g. logic programming),
but in the end, because it needs to be precise enough, it ends up
constraining the "autogenerated implementation" rather tightly.

Coding in C or Scheme is "writing a spec", and passing it to GCC or
Guile is "automatic generation of an implementation".


        Stefan