Chris Hanson wrote:
> a particular Scheme workshop at which there was discussion
> about the sad state of Scheme standardization.
Today, however, Scheme standardization is in fantastically great
shape, the R6RS process notwithstanding.
SRFIs are very effective as standards documents. When
implementors advertise their work, frequently they highlight
which SRFIs they've chosen to support.
Package systems (e.g., Eggs, Snowfort) are gaining momentum as
the framework for maintaining a shared body of "standard code".
The existing revised reports function well as reference points
in relationship to which implementors position their work (e.g.,
"full R5" or "R4, except for ...").
With just those tools, the community is, to belabor the phrase a
little bit, "working side by side" to grow the language.
What will R6 bring to the table? What will it bring to the
table that could not more effectively be contributed by
publishing SRFIs?
Interestingly, there is another language whose community is
organized as an open forum to which anyone may propose
standards, and a self-appointed "high authority" that blesses an
ever-growing set of standards documents as "the language": I
mean the Java community, of course.
With the SRFI system, and now the package systems, the Scheme
community has spontaneously, and just in the past couple of
years, really, built up its own version of a Java-like community
process -- but for one difference. Unlike Java, there hasn't
been (aside from the R6 process) any central authority whose job
is to bless the (ever growing) "official language". In some
sense, the R6 process tries to play that role but, at that role,
it is a terrible performer, with years between revisions and
very constrained communication between authorities and the
larger community. Meanwhile, people are growing the language
just fine, on their own, with SRFIs and package systems and an
abundance of implementations. Perhaps, in Scheme, we don't
really need that central authority any more.
The Scheme community has its "natural" authority figures:
implementors. On any given day, the Scheme family of languages
is, by definition whatever the implementors have provided.
SRFIs give implementors a way to say, one to the others, "hey
fellow Schemers, I think you'll want to try this". R6 would
say, instead, "hey fellow Schemers, if you don't do as I say,
I'm no longer sure you're really a Schemer". R6 -- almost any
R6 with large changes -- would put the community to "working
head to head".
That's not just theory, it's fact. It's already happening even
though R6 hasn't been ratified. It's happening right here in
this thread between Cowan and I. This is a complete waste of
time fight over a document title that happens to be a piece of
economically valuable real estate. Yet, waste of time or no,
it's also a necessary fight exactly because of the economic
value of that document title.
Bleh. I'm all riled up now.
Well, here's another compromise proposal since Cowan was only
mildly dismayed by the last one:
1) The editors vote tomorrow to shut down the R6 process, keep
this mailing list open, and convert to publishing a bunch of
SRFIs.
2) Everyone who is not an editor thinks highly of them for
taking such a noble (and "no bull") step.
-t
Received on Fri Jun 08 2007 - 15:43:30 UTC
This archive was generated by hypermail 2.3.0
: Wed Oct 23 2024 - 09:15:01 UTC