defsystem [was: Contributed code]

[moeller@informatik.uni-hamburg.de (Ralf Moeller)]

Daniel LaLiberte wrote:

>I suggest we start by using the defsystem from the lisp-utilities
>repository.  It is portable (although I had problems running it under MCL
>2.0p1, anyone got it to work?).  Cartier's defunit is very nice indeed, but
>unfortunately, it's not portable.  It would be good if Cartier's code could
>be added to the standard defsystem.  No need to implement everything for
>all platforms, just make it compatable.  So everyone could use the standard
>defsystem and mcl users would simply get more stuff.

Mark Kantrowitz wrote:

There is a proposal, by Kent Pitman, but there were several problems
with it, so it was rejected. (One of the key problems with it from my
point of view is that it assumes a procedural defsystem instead of a
structural defsystem.) If I ever get enough free time, I'll try
merging Kent's proposal with a structural proposal.

------------------
A few comments about defsystem:

Some time ago I followed a discussion in comp.lang.dylan. In Dylan
source code may not be stored in conventional files.
However, the current developments concerning defsystem for CL are inherently
file-oriented. Maybe we should keep in mind
that defsystem should also incorporate a non-file-oriented source code
management system.

In other words: A file defines an implicit order on the toplevel definitions
found in the file (call it (:serial <toplevel-expr_1> ... <toplevel-expr_n>)).
Furthermore, a certain compile-time environment is set up
for these toplevel forms! 

Any comments?

rm