RE: [DISC] Escaping #$#

[Roger Crew <rfc@MICROSOFT.com>]

> > I still want it in the spec though.
> 
> I don't want OOB quoting specified in MCP 2.1.  We don't have any 
> experience with this and there isn't an available server that 
> contains the solution we pulled out of a hat in the last month.
>
> I don't want to block correct MCP implementations on a new MOO server 
> release given that sweeping this issue under the rug has not, for the 
> most part, bitten any actual implementations.

Various points in no particular order:

(1) Blocking a specification of quoting on "experience" pretty much
guarantees that there will NEVER BE experience (... for now, I'll skip
the whole question of how many years from now MCP 2.2 might be coming
out...).  In order to gain experience, we have to have people building
clients and servers with quoting and they're not going to do that if
it's not mentioned or if there's only a vague note to the effect that
something needs to be done.

(2a) Saying this has not, for the most part, bitten any actual
implementations is a bit misleading.  You never really *know* when
you're getting bitten, the #$#-as-data lines just get quietly dropped on
the floor and you never hear from them again.  Even if it's only 1 line
in 10,000, there's no telling how important that line might have been.

(2b) If sweeping this issue under the rug really hasn't been a problem,
then it should likewise not be a problem for quoting to be in the spec
and for there to exist --- at least until such time as conformant MOO
(or other kind of) server becomes available --- implementations that are
conformant except in this one matter.

(3) There's plenty of precedent for releasing standards prior to the
existence of a totally conformant implementation.  In fact, this is
pretty much the norm, seeing as unless there's somebody on the committee
who's managed to WIN on every last issue, everyone with an
implementation will have work to do to get his/her implementation
conformant.
(we'll leave as exercises the questions of whether there exists a
totally conformant implementation of ANSI C, 8 years after its approval,
or whether there will *ever* exist a totally conformant implementation
of ANSI C++.  Not that this is something we should aspire to, but...)

(4) There's also some precedent for implementers ignoring stuff that
turns out to be braindead (e.g., trigraphs in ANSI C); thus far, no one
has offered any reason to believe the quoting proposal falls into this
category.
 
(5) If you really don't want to be waiting for MOO 1.9.0 (a not
unreasonable concern), there's my mostly-trivial patch that adds OOB
quoting to 1.8.0p6.  You could have a conformant MOO server *tomorrow*
if you really care...


At the VERY LEAST, MCP 2.1 should contain a *specific* proposal.  If
you'd prefer not to make it completely mandatory (e.g., as with macros
in R4RS), I can live with that, however a more substantive objection
would be useful.