Keith Moore <moore(_at_)cs(_dot_)utk(_dot_)edu> writes:
To turn this discussion into something potentially useful; do you
think a 2822bis (822tres?) should discuss X-* or not? If so, what
do> you think it should say?
If 2822 is revised I think it should only contain minor
clarifications of the existing material. The effort to revise 2822
should avoid opening up new cans of worms.
I agree. Do you also, like me, consider 822 existing material? 822
discuss X-*, and I think that clarifying its role in a 2822 revision
would be useful.
I consider X-* a can of worms.
Clarifying the X-* situation might be useful, but it shouldn't be coupled
with revising 2822.
The notion that people should use X-* for experiments (be them
implementation experiments or specification experiments) is quite wide
spread. Changing that notion, when there is an opportunity, seems
like a good thing.
It's perfectly fine to use X-* for experiments. What's not fine is
expecting to widely deploy and/or standardize the protocol used in those
experiments without changing it. If you want to widely deploy a field,
you need to formally define it, publish its definition, and probably
get some sort of community approval for use of that field.
There's really no problem with X-* at all; the rules in 822 are clear.
The problem is with people who think it's okay to widely deploy and/or
standardize half-baked ideas without doing either analysis or review of
those ideas. Changing how X-* is used won't address that problem.