Tue Mar 25 08:44:07 EDT 2008
maybe i should just stick to using an opaque representation, and try
to use this to gain access to the data.
ther are 2 models to use this:
* all code is available in source form: intermediate representation
doesn't matter much, since it can be redefined.
* some code is closed. in that case, the representation does matter,
because it becomes an interface.
it's the latter problem i'm trying to solve.
if this could be almost human readable, but mostly independent of
mzscheme's binary representation, that would be great.
* a dictionary needs to contain binary code + macro code + reference
to the compiler version / library it was made with.
* a dictionary needs to be opaque, and at the same level of scheme
what about separating the incremental model from the libary model? the
incremental model is mostly for developing. they are two different
* kernel: uses mzscheme's module name management system.
* incremental: extends the flat public interface.
maybe the next question is: what is a forth file?
how to make "forth file" == "scheme module"
and build the incremental compiler on top of that. it's very
straightforward for macros. but what about words? every module that's
compiled contains forth words the same way as the nested let.
yes. it's better this way.