[plt-scheme] Re: to define, or to let (last try)

From: Matthias Felleisen (matthias at ccs.neu.edu)
Date: Thu Apr 29 07:58:28 EDT 2004

I will remind everyone about one important thing: errors matter.
I will not elaborate on this issue. Instead, I want people to
think about this motto, because from the little that I read,
I could tell that people were not considering this.

Warning: my mail client does threads and I will delete messages
from this thread again.

-- Matthias

On Apr 28, 2004, at 11:42 PM, Bill Richter wrote:

>    Matthias Felleisen <matthias at ccs.neu.edu> responded to me:
>    As you all know, PLT Scheme has a fix evaluation order. We believe
>    it is the correct thing to do and until we see further evidence, we
>    stick with it. Evidence will *not* come with theoretical arguments.
>    It will have to come in new practical arguments. And believe it or
>    not: Matthew and I and everyone else of core PLT has seen a lot of
>    arguments and evidence on this topic. You will have to think hard
>    to create new evidence.
>    P.S. Good enough for a pronouncement?
> That's great, Matthias, but you're a top theoreticians.  Can you make
> a theoretical pronouncement about these 3 issues, and then we'll quit:
> 1) Definite semantics is important for reasoning about programs.
> 2) Sequential programming style is so bad that we want unenforceable
> constructions indicating coder's belief in eval order independence.
> 3) Depending on Mzscheme's left->right eval order creates a problem
> which didn't exist in non-buggy ambiguous eval order R5RS programs:
> (.... EXP ...) isn't equivalent to (let ([x EXP]) (... x ... ))

Posted on the users mailing list.