[r6rs-discuss] Implementors' intentions concerning R6RS

From: Elf <elf>
Date: Sun, 28 Oct 2007 05:24:41 -0700 (PDT)

On Fri, 26 Oct 2007, Thomas Lord wrote:

> Elf wrote:
>> i was not aware that 'a variety of scheme implemementations' meant 'my
>> implementation and two others, and to hell with what the rest of you
>> think'.
>> at what point does implementor and developer opinion matter? at what point
>> are we part of the entire Scheme community? apparently we're not right
>> now.
>>
>
> I think that that's exaggerated.
> The small number of implementations taking R6 seriously (and
> I would include Larceny!) are, also, the most actively maintained
> and furthered, fairly mature and serious implementations going.

how are we defining 'most actively maintained' and 'serious'? most srfi's?
largest user bases? most frequent version-control checkins? largest support
library set? most uptake outside the scheme community?

guile is/was used in tens of non-scheme products, including such high visibility
programs as gimp. can you say that of plt? chicken has been taken up by
swig (cross-language compatability), supports more srfis than any other
implementation, and has a new micro release every few days. slib is used by
all 'serious' implementations, and aubrey jaffer has already made his feelings
known about r6. srfi-1 and srfi-13 are supported by ALL implementations, yet
are not supported in r6.

chez-scheme has low uptake by most of the community because its commercial
(it is the only implementation i have not used). larceny only recently went
from a closed research project to something active. plt is generally a
love-or-hate implementation.

i would be very surprised if these three implementations together had a total
of 50% of the entire user base. speaking as someone who has taught scheme,
at least one of these implementations is the reason for most undergraduates
loathing and despising scheme. (i usually start people off with guile nowadays,
despite its non-r5 compliance, as the wizard book is still around r4 material.)

regardless of this, if the scheme community has always followed a consensus
model, how did three implementations get self-elected to represent the
entire community's wishes, and has it done a proper job of doing so?

the answer to the first bit is inherent in the question. the answer to the
second bit, if only those two-point-five (larceny being a point-five on the
support) implementations intend to support scheme, is a resounding no.
this is further evidenced by such projects as ERR5RS (which has already
gotten more uptake and acceptance than r6rs). the debate over the validity
or relevance of r6 seems confined to plt, chez, and a few sad buggers like
me who hate to see things going this way. its already failed.

>
> It's not so ridiculous to let those voices carry a lot of weight.
>
> Once again, the net effect (if all goes well) is that there will various
> approximations of R6 out there, and (if all goes well) libraries that
> are portable-in-fact will be developed, and (if all goes well) it won't
> be hard to adjust those libraries in response to further development.
>

who will develop these? plt and chez (and larceny). what about everyone
else? heres a fun question: find which implementations have the most code
written for them before determining who gets the most weight. ive done this
already. the voices you are saying should carry the most weight in general
should not.

-elf
Received on Sun Oct 28 2007 - 08:24:41 UTC