Ian G wrote:
As we haven't done much to add algorithms, I'd say this is fairly open
territory.
...and it's been handled in various ways by various other groups,
sometimes badly... certainly not consistently.... if an AD chair
wanted to comment on this that'd be nice. (And, if I'm missing
some process that's already there, my apologies.)
On the other hand, we probably want some minimal proof of need. By way
of example only, 2 cooperating implementations using an experimental
number?
The reason I say that is that we would then have 2 groups that would
also cooperate to write the specification, rather than one group that
asserts "it works for me!"
I agree that some level of consensus would be a good thing. I don't
know if there's a current way to formalize this in the IETF. In the
old days they allegedly required "three genetically separate
implementations". I see your idea as being along this line. Some
number of implementations GREATER THAN one.