[r6rs-discuss] reflecting on compilation

From: Thomas Lord <lord>
Date: Tue, 19 Jun 2007 15:59:31 -0700

What with all this crazy talk about portable solutions for
separately compilable library modules, how's about we
close the reflective loop on that a bit?

In particular, though it would require a delicate touch
to handle parameter setting and file-system issues well,
how about a portable interface for compiling a source
form and storing the resulting binary?

The aim of such an exercise would be to make front-ends
to the compilation process at least somewhat portable --
people could write *generic* front ends to compilation.

One advantage of generic front ends to compilation is
its application in portable package systems. The perl
community has done something like this for years, with
incredible success. It is a vital, technical cornerstone for
how perl has succeeded (to the degree it has) at "growing"
as a language.

Perl got a bunch of things wrong -- thing's I'd expect
Schemers to do a better job at. For example, typical
perl packages that involve non-trivial amounts of compilation
are a bit particular in their demands upon the host environment
and on where the resulting software should be installed.

I think Schemers would likely, intuitively, abstract that stuff
away and opt for a simpler model which is more easily
ported even among extremely diverse host environments.
A small extra burden on implementors -- to provide an implementation
of the abstract API for compilation in exchange for a big
payoff in the power of the package system.

Another advantage is the possibility for implementation-portable
Scheme IDEs. Reflecting on compilation in practical ways
is one cornerstone of how a code-base like the PLT Mr and Mrs
series of tools might become, gasp, *portable*.

-t
Received on Tue Jun 19 2007 - 18:59:31 UTC