[R6RS] Re: Less verbose type definition form

Michael Sperber sperber
Tue Jul 19 12:56:36 EDT 2005


feeley at iro.umontreal.ca writes:

> I'm sorry, but if it takes more than one line of code to define a 2D
> point record, then the record system is not elegant.

That's your notion of elegance---I just don't agree with it.  Which is
exactly what makes this kind discussion unlikely to lead to agreement.
The record-type definitions I want to seem distinctly awkward in your
syntax.  (By some subjective notion of awkwardness, of course.)

"One-line record-type definitions" are a reasonable thing to ask for,
but Kent *did* post a suggestion to enable exactly that, which you
haven't responded to.  (Along with a rationale of why he'd prefer not
to do it.)  Generally, we've tried to accomodate what people
want---implicit naming, inheritance, final classes, opacity, all way
beyond the requirements.  I specifically have gone (I feel) out of my
way to accomodate wishes I fundamentally disagree with.

> Moreover, I view "extensibility" as such a strong requirement that
> it is a design principle.  This is why I believe a syntax based on
> keywords, like the one I proposed, is necessary.

You haven't shown that Kent's syntax is not extensible---it is also
keyworded.  I believe Kent has a conservative extension to the
proposal for a full-blown class system.  Which extension specifically
do you think is made impossible by the draft?

> What I don't understand about your comment is that you say we will never
> agree on the syntax so any syntax will have to do and then when I say
> syntax is very important and the syntax I propose is better (succinctness,
> readability and extensibility while having the same expressive power as
> your proposal) you seem to not want to discuss it.  I just don't
> understand what the problem is.

Kent and I both sent in responses to your syntax.  We disagree that
"your syntax is better."  You claim "elegance, "extensibility," and
"better" for your proposal, as though these were obvious axioms.  It
seems you don't believe me when I say I don't think your syntax is
better.  I acknowledge that you do believe yours is better, and
there's probably no way I'll convince you otherwise.

My own ideal syntax would look different still---because, for myself,
I can cater to different goals.  Catering to all of them at the same
time is impossible.  (Or at least none of the proposals advanced so
far does it even remotely.)  Just giving you all of what you want and
giving me all of what I want at the same time is impossible, because
our goals conflict.  Which is why I haven given up on some of the
things I want for the purpose of reaching agreement.

All in all, the thing to do is to let those people who like the record
syntax use it, and for everyone else to build their own on top of what
we provide.

There are some comments about your syntax that I haven't posted
yet---if you insist, I'll send them in.  But again, this seems an
ultimately pointless exercise for the reasons outlined here and in my
previous posts.

-- 
Cheers =8-} Mike
Friede, V?lkerverst?ndigung und ?berhaupt blabla


More information about the R6RS mailing list