Cross-Area Review (was: Fwd: RE: A follow up question onietf@ietf.org)

[James Kempf]

Brian,

I've found the best vehicle for cross-area review to be directorates,
but they are very informal and their organization and use is very
dependent on AD energy. Since ADs have a lot to do, especially if they
have demanding day jobs in addition to their IETF responsibilities,
one's milage with directorates does vary.

The primary issue is getting cross area review of work in progress.
Most WG chairs don't have the cycles to follow other WGs work in
detail, so directorates provide a forum where WG chairs can informally
discuss work in progress and get information when issues arise that
need attention.

The other vehicles you've mentioned below (charter reviews, IAB
attention) involve people who are not directly involved in the work,
and so it is often difficult for them to follow the work in detail. By
the time an issue is raised to the level of a charter review or comes
to the attention of the IAB, it can be larger scale.

            jak

----- Original Message -----
From: "Brian E Carpenter" <brian at hursley.ibm.com>
To: <problem-statement at alvestrand.no>
Sent: Friday, April 25, 2003 1:24 AM
Subject: Re: Cross-Area Review (was: Fwd: RE: A follow up question
onietf at ietf.org)


> One of the questions here is why the mechanisms that we have today
> don't always work.
>
> We have open review of WG charters, where cross-area issues
> really ought to be identifiable.
>
> We have the IAB, which keeps an eye on all BOFs and is supposed to
> have a cross-area mandate in general.
>
> We have IAB workshops, which are definitely supposed to take a wide
> view.
>
> Are we lacking something a bit later in the process, after a WG
effort
> has been started but before its results have crystallized?
>
>    Brian
>
> Keith Moore wrote:
> >
> > >  I've seen a number of postings that are coming
> > > close to saying "we approved IPv6 site locals as part of a
> > > proposed standard in RFC 1884, and again in RFC 2373, and again
> > > in RFC 3515, without enough application community review to
> > > shake out the issues we're shaking out now on the IETF
> > > discussion list".
> > ...
> >
> > > But I am curious what the lesson might be, for those of us
> > > thinking about the problems with IETF standards process.
> >
> > well, rather than extrapolating from one data point, maybe we
should consider
> > some other snafus also.
> >
> > I'd argue that IPsec largely missed the boat by sticking to a view
that hosts,
> > and in particular IP addreses, were generally appropriate to use
as security
> > principals.  As a result applications still don't use it today.
> >
> > lots of work has been invested in service location stuff which for
some reason
> > has never resulted in much use.
> >
> > URNs (one of my pet projects) don't seem to have been as useful as
> > anticipated.
> >
> > The zeroconf work has produced multiple disasters, including both
v4 linklocal
> > addressing and LLMNR (which was spun off to the DNS folks but
still suffers
> > from flawed assumptions IMHO).
> >
> > --
> >
> > one kind of problem seems to occur when there is a long delay
between the time
> > that the fundamental design decisions are made and the time that
they actually
> > start to get used.  a tighter feedback loop between specification
authors and
> > early adopters might help.
>