gjc@mitech.COM (03/11/91)
The lack of SQUID and the foolishness of the "#," approach first struck me when dealing with the port of the FORTRAN->LISP translator from MACLISP to COMMON-LISP. By "#," approach I mean a tendency, first seen in the LISPMACHINE at MIT, to avoid an *OPEN* implementation policy, generally a SEMANTICS FIRST, then SYNTAX, has a tendency to be the most open, and instead go as soon as possible to some concept limited by the imagination of the lisp implementor of *EXACTLY* what the user will *REQUIRE*. One result is that many significantly complex applications, e.g. Macsyma, Fortran->Lisp, generally any imbedded language, depended critically on internal undocumented representations (e.g. the readmacro expansion of "#," and internal flags such as SI:COMPILER-XYX-FLAG) so as to be frustratingly fragile to systems changes. (By FORTRAN->LISP I mean the one done at MIT in the Mathlab group by KMP). Does there exist in the Scheme community at least an implementation effort which is not an end in itself? -gjc