# Date: 18 Mar 1996 15:39:06 GMT
# From: sjenkins@iastate.edu (Steven L Jenkins)
#
# I'm seriously contemplating porting scsh to mzscheme (the Scheme
implementation
# from the PLT group at Rice)...I've talked a bit with Shriram about this and
# have done some preliminary poking. While I'm not exactly thrilled over the
# idea of reimplementing scsh, it seems my only option if I want to get decent
# performance (as well as a system that will run on an Alpha). Does anyone
# (including Olin & Brian) have any tips for this? Would it be more
# realistic to build my own implementation of PreScheme that is portable than
# it would be to add scsh to mzscheme?
As daunting as porting scsh to another Scheme system sounds, I think
it pales in comparison to hacking PreScheme.
If *I* were to start porting scsh to another Scheme system, I would
take the approach of using the manual as a functional specification
and delving below the surface into the source code only when it seems
like it will save a bunch of coding. For example, it may may make
sense to implement a compatible macro system in order to re-use the
hairy process-oriented macros---on the other hand, it might make more
sense to implement the macros as needed using the native macro
system.
This top-down approach would allow me to implement the pieces I needed
more quickly, and would allow me to see immediate progress, rather
than spending a few months porting low level things that have little
relation to scsh itself.
The alternate approach of creating an infastructure which allows the
scsh sources to compile/run without modification on another Scheme
system is, in my opinion, a really tough job that would result in a
pile of hard-to-maintain comptability code. On the other hand, you
probably wouldn't have to maintain the scsh code itself---you would be
sharing it with the Scheme 48 version---so it might not be so bad
after all.
Brent
|