On Thu, Oct 02, 2014 at 01:31:05PM -0700, Bob Braden wrote:
> Possible properties of protocol names might be:
It should be possible to do this as aliases of numbered RFCs.
In some areas we already do this informally.
One of the nice things is that some aliases could be changed as RFCs get
obsoleted. Just like STDs.
> (1) names, not numbers!!
> (2) Probably hierarchical; think domain names.
Whatever's appropriate, sure. Hierarchical naming won't work for some
things that have something like multiple inheritance.
For example, would RFC4121 be Kerberos.GSS or GSS.Kerberos??
Also, we already tend to speak of GSS-xyz when we mean "a GSS mechanism
based on xyz", and "xyz-GSS" when we mean that protocol "xyz using GSS".
I'd rather not make hierarchical naming a hard requirement, just give
guidance and let those picking the names do so wisely.
> (3) Might incorporate the "Obsoletes" attribute by some sort of
> generation number.
> ("TCP.1", "TCP.2," ...)
Hmmm, maybe. Combining (2) and (3) we might end up with names like
PAWS.1.TCP.2 or PAWS.1:TCP.2, ... Eww.
I'd rather that the aliases follow the latest, then we'd have just
PAWS.TCP. But it's true that being able to refer to a past version of
something would be convenient.
This is worth doing. I'm not sure that we need to do massive surgery on
the IETF process here. I'd just allow something like first-come-first-
served assignment -by WG chairs and IESG members- of aliases to RFCs.
Later on we might require this for new publications.
Later still we might deprecate numbering for new publications.
KISS and all that.
Nico
--