Re: Please rename TRY to something else Manuel Serrano (29 Aug 2002 07:23 UTC)
Re: Please rename TRY to something else sperber@xxxxxx (29 Aug 2002 09:02 UTC)

Re: Please rename TRY to something else sperber@xxxxxx 29 Aug 2002 09:02 UTC

>>>>> "Manuel" == Manuel Serrano <xxxxxx@sophia.inria.fr> writes:

Manuel> [...] was against some of the first Olin's SRFI
Manuel> (at least SRFI-1 I think) that was overriding functions definitions
Manuel> of Scheme R5Rs. Now, it is not question to override a Scheme definition but
Manuel> it is question to override a Bigloo definition. May be it is only a concern
Manuel> for Bigloo users but it is a real one.

I see your problem, but I don't know that there's a way to avoid it in
general:  Sure, Olin could have avoided re-defining R5RS-bound
names---R5RS is a published specification.  But how can you avoid
choosing a name chosen by the author of some Scheme implementation
that's out there?  There's no formal registry, there are dozens and
dozens of them, and new ones are sprouting about every other month.
Of course, Bigloo is more mature and more high-profile than most
others, but that's beside the point in this context.

Moreover, Scheme implementors are as likely to choose intuitive names
as SRFI authors, as with TRY.  Unfortunately, the whole naming issue
with exception systems is a veritable mine field as there are so many
names to choose from or to avoid, depending how you look at it.
Carrying through your argument means that any Scheme implementor can
effectively hijack any names she wants and argue against a SRFI (or
some other library mechanism) choosing that name as well.
(Technically, I'd say that a module system should be able to deal with
this kind of problem easily, but that's again not the main point
here.)

Manuel> I don't think that it is possible to change Bigloo to add a new
Manuel> semantics for try. The problem is not of how to implement it (it could
Manuel> feasible). The problem is that it is a bad thing to have two different
Manuel> semantics for one unique form. I have hated so long the Caml syntax
Manuel> that used to provide two forms from writing the same expression (LET
Manuel> vs WHERE), not to add now two semantics with one unique name in Bigloo
Manuel> :-) I see only bad reasons for trying to do so.

Richard and I are fully willing to change the name if we have an
alternative that we agree with.  He hasn't argued against changing the
name in principle, he only argued against TRY/CATCH.  So I'd say keep
the suggestions coming.

--
Cheers =8-} Mike
Friede, Völkerverständigung und überhaupt blabla