Document Blocking (Was: I-D ACTION:draft-ietf-problem-process-00.txt)

[Thomas Narten]

Going back through this thread to understand better...

Keith Moore <moore at cs.utk.edu> writes:

> > >But if one or both Security ADs are deeply convinced that a draft 
> > >constitutes a major security risk, or one or both Routing ADs are 
> > >convinced that a draft will lead to routing loops, isn't it quite 
> > >appropriate for them to block the document?
> > 
> > Absolutely not, but *not* because the document shouldn't be stopped. 
> > The ADs who think that there is a serious problem with a document 
> > should convince the rest of the IESG that the document is a bad idea. 
> > Then, the IESG can come to consensus (or unanimity) to reject a 
> > document (or at least stop it until the problem is fixed).
> > 
> > The problem with the current process (as I understand it) is that it 
> > allows documents to be blocked by one or two IESG members without the 
> > consensus of the group. 

> I don't see this as a problem at all.

I see a problem with Keith's statement here. It can be read to imply
that Keith thinks it's fine for an AD to block a document without
there being any other support for the action by the rest of the
IESG. I disagree with this view, and I doubt anyone else on the IESG
would agree with this view. I hope that I'm misreading Keith's intent
here.

What is the case is that there is an assumption that when an AD raises
a discuss, there is validity behind it. Whenever there is a discuss on
a document, the issue needs to be articulated so that the rest of the
IESG and the IETF community as a whole can be made to understand what
the issue is. If that can't be done, we have a fairly basic
problem. Our internal processes call for all discuss's to be written
up. Once one has words, one can discuss them and agree/disagree, claim
lack of understanding, etc. But the notion that an AD can (or does)
just put in a vague/bogus discuss and no one ever objects to it is
broken shouldn't be happenning. If it is happening, that's broken, and
needs to be fixed. Again, here, having specific examples to point to
would help. (I'll grant some people are saying there is an appearance
of this happening, and our process rules don't prevent it from being
possible, and therefore this aspect of our processes needs to be
changed.)

Now, you can read my words above differently depending on whether you
think the cup is half empty or half full. I.e., maybe the IESG just
defers to others and doesn't even check whether a particular discuss
has any validity. Or, that all IESG members review *every* discuss in
detail and agree with it. The reality is somewhere in between. For
example, when the security ADs raise an issue, I do review them, and I
have found that in most cases I agree with them. So I don't feel like
I need to explicitely check every thing they say -- they have a track
record and I have found them to be reasonable.

> Many of the issues that hold up document publication are not easily
> understood without expertise in that particular area.

But it is also an absolute requirement that a real issue must be
explainable to the community. So suggesting that "the issue is too
hard to understand by mere mortals and therefore an AD must have
blocking power" is unnacceptable.  IMO, the "one discuss blocks" rule
isn't designed to allow the above, but is intended to allow any AD to
raise an issue that can't simply be dismissed, say, because only one
AD voted that way. 

> Also, IESG does not "block" documents, it explains what is wrong
> with them and sends them back to the working group.  It's simply
> infeasible to expect all of IESG to reach consensus on every issue
> that requires a change to a document. For a deep technical issue,
> there *might* be four people on IESG who really have enough
> appreciation for that issue to express an informed opinion - and two
> of those might have to stretch to understand it.

I would couch this differently. Every AD must be capable of being able
to understand an issue being raised. This is part of our need for
breadth. But it may take time -- or even a significant amount of time
-- to do that in cases where the issue is subtle, etc. Requiring that
all ADs understand everything to this level is simply not
scalable. But the other extreme of one AD "understanding" an issue,
but no one else (either in the WG or on the IESG) is also broken.

Thomas