On Fri, 5 Nov 2004, Nikodemus Siivola wrote:
> On Fri, 5 Nov 2004, Zbyszek Jurkiewicz wrote:
>
> > I agree, but then you must decide where is the boundary between what is
> > feasible and what is not --- and the decision depends on current
> > technology state (we should only define the behaviors we know we can
> > implement). The agreement in this case should come from language standard
> > group or similar body, otherwise we end up with nonportable programs.
>
> This is rapidly going OT, but... the historical precedent for CL is to
> standardize on existing practise, not to implement blue-sky standards.
>
> Existing practise grows because people extend it, just like Thomas is
> trying. Critising an extension plan because there is no standard for it is
> silly, even though critisising it for not playing nice with standard
> facilities may be quite valid.
>
OK, let's make it clear, I am not "critisising" it (by the way, standard
come when people are trying to agree on something, not when some
important commitee puts its placet).
I am rather saying that if we want add new construct we have to provide
it with sufficient semantics. Remember PROG is still with us, and it is
typical construct added in completely ad-hoc manner to the language.
In the case of dynamic-wind, there is/was
some discussion about it between Scheme implementors and users.
Personally I do not think dynamic-wind fits Common Lisp, maybe we can
find some better construct. For example, dynamic-wind does not allow you
to bind anything, thus if you open the file, you must put dynamic-wind
inside external let and use setq in its prologue while opening file.
Regards,
Zbyszek
