Re: Cross-area review (was Meeting rotation)

[Brian E Carpenter <[e-mail]>]

On 24/12/2015 00:17, tom p. wrote:
> ----- Original Message -----
> From: "Brian E Carpenter" <[e-mail]>
> To: "Andrew Sullivan" <[e-mail]>; <[e-mail]>
> Sent: Saturday, December 19, 2015 12:02 AM
> 
>> On 19/12/2015 12:36, Andrew Sullivan wrote:
>> <snip>
>>> Finally, there's been a great deal of worry about the lack of
>>> cross-area review that we get these days,...
>>
>> I'd want objective evidence that we get less cross-area review than
>> at some date in the past before worrying a great deal. It seems to me
>> that we are actually doing it more systematically than in the past.
> 
> Brian
> 
> I think that we are doing more, in the shape of Directorate, Gen-ART and
> such like reviews but that they may not make the I-D better.
> 
> Any one operating any sort of system will likely know that change
> introduces faults which introduce change which ....   The art is to get
> it right first time and when these late stage reviews introduce changes,

>From what I've seen, mainly as a GEN-ART reviewer, *changes* are not what
happens at that stage. Corrections - yes. Clarifications to remove ambiguity
- yes. But changes of the intent of the protocol, or changes in choices that
were a matter of taste or WG preference - no.

The IESG is also explicitly supposed to avoid changes for reasons of taste
or preference:
http://www.ietf.org/iesg/statement/discuss-criteria.html#stand-undisc

> then I see the I-Ds getting worse, certainly in terms of the coherence
> and consistency of the language, probably in terms of technical accuracy
> too.  Outside people suggest new text which makes sense within their
> mindset but may not in the context of what the WG has so far produced;
> and by this stage, the WG's energy may be waning, so something that may
> have been corrected in a -00 or -01 gets left in in the I-D-(RFC minus one).

Yes, that can happen, depending on the shepherd and authors. But it mainly
happens in the WG I believe, during the long grind to achieve a WG Last Call
with no dissenters.

> The best I-D, in some senses, is often the first to be produced, before
> adoption by a WG, but that almost always has to be changed and the
> longer that that goes on, commonly years, then the less coherent the I-D
> is.  I am thinking particularly of I-Ds that are too large to be held in
> one's mind all at once (which most are).

Sure. And that is where another point comes in: the running code point.
I watched the DNCP/HNCP sausage being made. Guess what? Those are two complex
documents but most of the later changes were in direct response to cross-area
review comments *and* comments from implementors. The most powerful response
to a review comment is one based on implementation experience.

    Brian

> 
> Tom Petch
> 
>>> ...  but it doesn't seem to me
>>> that the meetings obviously help with that except by accident.
>>
>> But those accidents - hearing about something by chance that you would
>> never discover by chance on-line - are a major part of the advantage
> of
>> our plenary weeks. Being all in one place and time zone allows those
>> useful accidents to happen.
>>
>>    Brian
>>
> 
>