At 11:13 12-06-2012, Brian E Carpenter wrote:
I don't understand that statement. RFC 5226 says, in Section 2 about
"Why Management of a Namespace May Be Necessary":
" A third, and perhaps most important, consideration concerns potential
impact on the interoperability of unreviewed extensions."
One of the specific considerations for designated experts in section 3.3
is
" - the extension would cause problems with existing deployed
systems."
It seems clear that interoperability is a primary concern for any
expert review.
This gets you into the level of review (see Barry's comment).
The easiest policy is FSFC. The next step is Expert Review which is
supposed to have clear guidelines. If you want interoperability,
there's Specification Required. It seems from the above that the
argument is about avoiding problems with deployed systems.
At 11:38 12-06-2012, John C Klensin wrote:
I've with Barry on this. The details of the expectations of an
expert reviewer, including the thresholds for approval, should
be specified in whatever document sets up the particular
registry. One size does not fit all; "Expert Review" is a
designation of a mechanism and not a set of criteria.
Agreed.
We should, IMO, do two things in this area:
(1) When a document specifies "Expert Review" for a registry, it
should be required to spell out the criteria the Expert is
supposed to use, at least to the degree that isn't obvious. If
it doesn't, that should be grounds for "DISCUSS until fixed".
I like the idea of spelling up the criteria.
(2) If it turns out that an Expert for a particular registry is
not behaving as people expect, part of the process for getting
that fixed (or even complaining about it), should be to see if
the registry-creating documents are clear about procedures and
criteria. If they are not, an effort to update those criteria
would be a useful way to discuss the issues and not the
individual expert. Of course, Experts who knowingly violate
clear criteria should be summarily fired -- but I think we can
trust that to the IESG and note that it has almost never been
necessary.
The people who come to the registry may not know much about the
process. I wonder if they are aware that there is a path to complain
if there are not satisfied. Even then, they may find it easier to
walk away. I don't think that the individual expert is the
issue. The better path is to have clear documentation (what you
mentioned above) about the criteria so that the registrant can tell
what the expectations are. I'll add publication of the median time
to process a request as information for the registrant.
Regards,
-sm