ietf
[Top] [All Lists]

Re: I-D Action: draft-wilde-updating-rfcs-00.txt

2016-12-11 18:39:35
On 12/12/2016 10:23, Scott O. Bradner wrote:
see, for example, 
https://datatracker.ietf.org/doc/draft-ietf-newtrk-repurposing-isd/

And while we're reviewing ancient history, let me say that the new IESG in 2005,
with me as the new Chair, did spend hours discussing that draft and failing to
reach a useful consensus. But not because we thought there was no problem. As 
I've
said more than once, there is a problem, for any protocol that is complicated 
enough
to need several interlocking RFCs to define it. As those various components 
require
updating, we grow a dependency tree. The "Updates" tag on the more recent RFCs 
is a
very coarse way of expressing the dependencies.

Requiring the updating RFC to be clear about why and how it is updating other 
RFCs
is IMHO a good idea. However, I don't think that a mandatory section in the 
updating
RFC is the right way to ensure this. It would just become a box-ticking 
exercise.

Regards
    Brian


Scott (network WG chair)


On Dec 11, 2016, at 2:22 PM, John C Klensin <john-ietf(_at_)jck(_dot_)com> 
wrote:

Erik,

Sorry for the delay in responding. Let me try a very high-level
summary of the implications of what I, at least, consider the
most important history of the problem you are trying to bite off
a piece of (others will have other histories).  First, it isn't
easy.  Even if one just ignores the various flavors of
Informational documents, the right documentation rules for
single-stage processes (e.g., BCPs) are inevitably different
from those for two (and previously three)-stage technical
standards track ones.  That problem is further complicated by
the fact that we use BCPs, and occasionally technical standards,
for what are really procedural or policy statement documents.
Second, there is a complexity tradeoff.  Today, for normative
documents, we have BCP documents and 2 1/2 levels of standards
track  one (depending on what one thinks Experimental is).  

The issues of updating and categories are also inevitably
complicated by the nearly-orthogonal one of interrelated and
interdependent documents, some developed at different times and
by different groups and often with non-obvious overlaps.

We tried to take all of this on some years ago in a WG called
"NEWTRK".  It was not successful.  In particular (and trying to
state this as neutrally as I can manage), the WG concluded that
we needed a new type of Standards Track document that would talk
about status and relationships among documents, rather than
being one or more technical specification itself.  At least some
of what you seem to be proposing would go into those
standards-description documents and not the technical
specifications themselves.  In addition, at least some of us
believe that the relevant documents would be living documents
with change histories rather than the inherently-static (at
least per-document) RFC series.  

WIthout revisiting the argument and various opinions about
motivations, the IESG concluded that the idea was just too
complicated and/or inappropriate and the idea when nowhere.  In
retrospect, they might have been right.  Or not.  Either way,
the experience left many of us reluctant to start nibbling at
the issues again unless there was a comprehensive plan that the
IETF was willing to sing off on.

However, I do believe that it is unrealistic to believe one can
take on inter-document relationships without at least reviewing
the issues that the NEWTRK WG examined and the options it
considered.

best,
   john



--On Wednesday, December 07, 2016 21:52 -0800 Erik Wilde
<erik(_dot_)wilde(_at_)dret(_dot_)net> wrote:

hello john.

On 2016-09-15 09:37, John C Klensin wrote:
Independent of where it is discussed (as
long as it is on a public list), this I-D would be, at least
IMO, a much more satisfactory basis for discussion if it
demonstrated more convincingly that the author was aware of
those earlier discussions and had considered them, rather than
assuming (or appearing to assume) that no one had thought
about these topics.

it was not my intention to ignore or belittle previous
discussions. it just occurred to me as a frequent reader of
RFCs that there is a large variation in quality how updates
are documented. the idea was that some simple documentation
guidelines might help to improve that, without necessarily
being hard definitions on what exactly updates are, and how
exactly they have to be documented.

i'd be more than happy to include these earlier discussions,
but i am afraid if that involves going through a long list of
mail archives, this simply is beyond the time commitment i can
reasonably make. i'd be more than happy to have somebody
co-authoring and filling in those gaps, but that of course
assumes that somebody else would be willing to put in the
effort of writing up this history.

thanks and cheers,

dret.