Michael,
Woah! I haven't seen you on this list before. Welcome!
I used to *hate* R6RS. This was of course, before using it. Now, my
favorite implementations are among the R6RSers... (sheephish grin). So,
it doesn't seem right to have gone this long, enjoying the fruits of the
R6RS standard without saying, thanks! You did a great job! :-)
Ed
Michael Sperber wrote:
Derick Eddington <[email protected]> writes:
[1] http://groups.google.com/group/ikarus-users/msg/8392014ed8902e4c
As the post points out, optimally resolving library constraints is hard.
That's because figuring out a correct link of sets of libraries from
different sources its hard: The R6RS constraint language merely allows
you to express that complexity. I agree that this complexity is
frustrating, but at this point, I see it pretty much as a fact of life.
It does this mostly because of my bad experience with more limited
schemes (greatly over schemes without versioning - look at the trouble
the Java people have, for instance), and I still prefer the R6RS scheme
to the others I've seen.
I didn't expect realistic Scheme implementations to implement an optimal
constraint solver. (BTW, Richard Kelsey's implementation of SRFI 7 -
which supports something similar - also has a limited solver in it.)
Rather, I expected either the implementation to make simplicistic
choices as it goes along, or the user who assembles a final product to
specify (via a "linker command line" or some other external mechanism)
what versions she actually wants to satisfy the constraints.
