[r6rs-discuss] Implementors' intentions concerning R6RS

From: Jon Wilson <j85wilson>
Date: Tue, 30 Oct 2007 08:21:48 -0400

Hi Aziz,

Abdulaziz Ghuloum wrote:
> To put it another way, if an implementor never planned on
> ever adding a full numeric tower, hygienic macros, call/cc,
> dynamic-wind, multiple values, etc etc to his Scheme
> implementation (call it a pre-R4RS era Scheme), then it's
> only expected that such implementor is not going to adopt
> R6RS. As a Schemer, I don't care about such implementation
> and its implementor's opinion about R6RS.
> It's simply irrelevant.
>

Or to put it yet another way, you feel that you can safely ignore the
opinion of those who disagree with you simply because they disagree with
you? Danger, Will Robinson!

>> Then why bother writing and ratifying R6RS?
>
> Because there are people out there who apparently do care
> about the stuff that's in R6RS. Again, if the implementor
> of your legacy Scheme implementation of choice never cared
> about R5RS, then sorry, but that does not mean that we all
> have to suffer the consequences. Those who bothered writing
> and ratifying R6RS care about seeing it happen, and it is
> sickening how their actions and motives have to be
> challenged at every occasion.

Actually, I'm concerned more about the large number of implementations
which did adopt R5, which also have no intention of adopting R6.

>
>> As I understood things, there were two goals for R6RS:
>>
>> * Fixing a number of minor issues and ill-defined things in R5RS
>> * Increasing the cross-implementation portability of Scheme code
>
> Site your sources. The first page of the ratified R6RS
> standard document lists a different set.

Cite my sources... well, my source for that statement was my overall
impression and understanding of the purposes of R6RS. Hence my
statement "As I understood things..." In some sense, I did cite my
sources. Perhaps I simply misunderstood things, then.

> And how do you propose that "increasing cross-implementation
> portability of Scheme code" is going to be achieved if your
> implementor rejects anything that goes beyond their current
> set of favorite R[345]RS picks?

By writing a standard which is good enough that the implementors of
"legacy schemes" will want to implement the new standard, and which also
does include a minimal module system for the sake of portability.
Perhaps this is impossible. But I have yet to see any evidence to that
effect.

> R6RS intends to "allow programmers to create and distribute
> substantial program and libraries (e.g. SRFIs) that run
> without modification in a variety of Scheme implementations".
> The "Scheme implementations" are those that adopt R6RS of
> course.

Slow down, cowboy! Actually, everyone check your watches. This is
first time I've seen this, but I doubt it will be the last. Now, R6RS
is "Scheme", and those who don't adopt R6RS are doing some other than
implementing "Scheme". Only Chez, Scheme48, and PLT are now "Scheme
implementations".

I'd rather you didn't hijack a historical name in order to
de-legitimatize those who disagree with you. I understand that there
are a lot of other Schemers, many of them _much_ more seasoned than I
am, who would agree with me.

> Now whether this will be achieved is yet to be seen,
> but I can see it taking shape already (see SRFI-41 and try
> loading the R5RS code in your favorite Scheme implementation).
>
>> Since the standard was such that very few implementors were going to
>> bother with it
>
> Seriously, do you think that the opinion of some of the
> implementors whose implementations have 0% user base have
> any weight as far as how everybody else should go about
> doing business?

Well, yes. I also think that the users of those implementations with
"0% user base" should have some weight. Clearly, the editors thought so
too (at some point at least), since they opened the vote to everyone who
registered. In the past, consensus was required, not just a large
majority. I suspect that this would have been a wise tradition to
continue. I also think that the opinion of people like Olin Shivers,
Jonathan Rees, Marc Feeley, William Clinger, Gerald Sussman, and others
ought to have some weight...

If the editors really only valued the opinion of the implementors of
PLT, Chez, and Scheme48, then they would have just asked those
implementors whether R6RS should be ratified. Your efforts to
marginalize the rest of us are becoming a little ridiculous.

>
>> ... the standard had already failed its most important goal!
>> ...
>> ... A standard which is largely unimplemented is a useless standard ...
>
> You forgot to mention putting your tin-foil hat on.

Ahhhhh, and the timeless ad hominem attack. One of my favorites. This
one has a nice texture, pleasantly crunchy, but lacks flavor and body.
Regards,
Jon
Received on Tue Oct 30 2007 - 08:21:48 UTC