[Old-standards] Re: [newtrk] Re: Historic (Re: List of Old Standards to be retired)

[John C Klensin]

--On Sunday, December 19, 2004 12:39 PM +0100 Harald Tveit 
Alvestrand <harald at alvestrand.no> wrote:

> --On lørdag, desember 18, 2004 11:51:56 -0800 Bob Braden
> <braden at ISI.EDU> wrote:
>
>>   *> >>
>>   *> > This must be some new redefinition of the meaning of a
>>   Historic RFC.   *> > In the past, it meant "don't do it
>> this way anymore, we no longer   *> > recommend it, there's
>> another way to accomplish the same goal".   *> > So, for the
>> PPP items listed, what's the better way to accomplish the
>> *> > same goal?
>>   *>
>>   *> No, it's the old definition of Historic.
>>   *>
>>
>> Harald,
>>
>> I am puzzled by your comment.  I believe that Bill Simpson is
>> correct about the "old" (historic) definition of Historic
>> category, defined by Jon Postel.  Jon believed that if you
>> have a standard defining interoperability, it is ALWAYS a
>> standard unless there is a compelling reason to warn people
>> away.  The IETF can change the meaning of Historic, but let's
>> not change history.
>>
>
> With all respect to Jon Postel - the IETF's meaning of
> Historic is defined by reference to IETF consensus, not to Jon
> Postel's opinion.
>
> We may be confused by different meanings of "old" - I was
> referring to 1994.
>
> Shows how young I am, I guess :-)

Harald, I don't think it is because of your tender age, but it 
seems to me that "defined by reference to IETF consensus" puts 
us into dangerous territory.  The justification for this 
de-crufting effort seems to me to be (i) finding a relatively 
lightweight way to get what we are actually doing aligned with 
the requirements of 2026 and (ii) begin more clear to relative 
"outsiders" about just what we intend and are encouraging.

That suggests two things to me...

(1) A definition, whether in 2026 or earlier, or developed by 
some newer IETF consensus process, is useful only if the 
definition itself is absolutely clear.  We aren't there any 
more, if ever we were.   The language in 2026 isn't clear enough 
to prevent several of us from disagreeing about what it really 
means.  There hasn't been a clear statement and a consensus call 
on it in the last decade or more.  We aren't making progress if 
our definition "by IETF consensus" requires a royal assertion of 
what that consensus is, without a clear basis in IETF 
consideration.

(2) Part of what has gotten us into this mess is that we got rid 
of the old "required"/ "recommended"/ "not recommended" 
categories that were orthogonal to standards maturity levels. 
Once upon a time, we could say "X is a standard because it is 
interoperable and widely deployed, but we have concluded that it 
stinks so no one should be implementing and relying on it". 
That is "standard, not recommended".    Too much of the 
discussions of the last few weeks feel to me like an attempt to 
conflate (or avoid conflating) "Historic" with "Not recommended" 
and that just doesn't work very well in the edge cases... and 
there are lots of edge cases.

   john