Re: [admin-discuss] Next steps towards a net zero IETF

[John C Klensin <[e-mail]>]

TL;DR summary: WG and AD accountability should be seen as a
tradeoff with AD workload, rather than treating one or the other
as absolute.  We (especially the IESG) also should be more
explicitly considering the amount of high-quality work the IETF
can do in parallel and potentially setting higher thresholds for
WG chartering and continuing as a way to keep workload under
control.  In several respects, we may be dealing with a quality
versus quantity tradoff and that should be much more explicit.

Details and reasoning inline below.

(written before Eliot's more recent note although being sent
after it was posted)


--On Wednesday, April 5, 2023 09:19 +0300 Lars Eggert
<[e-mail]> wrote:

> Hi,
> 
> On 4. Apr 2023, at 21:34, John C Klensin <[e-mail]>
> wrote:
>> So that takes us back to variations on the
>> suggestion Eliot made earlier and a question to the IESG: I
>> understand "wanting" more session time but do you have a clear
>> sense --ideally one on which you can report to the community
>> -- about rationale and need?  When those sessions occur, are
>> they about presentations, status reports on what the WG is
>> doing, or discussions of open issues?  Of the latter, what
>> fraction of them are issues that have been thoroughly
>> discussed on mailing lists but remain unresolved rather than,
>> e.g., ones that have been deferred to or saved for f2f
>> discussions.  And, for those WG's who have asked for and
>> gotten more than a one-hour slot, what is the evidence that
>> the additional time was actually marginally productive in
>> issue-resolving or problem-solving?
> 
> We've always trusted the WG chairs to make that determination,
> and use the different available participation venues (mailing
> list, in-person and remote interims, in-person meetings, etc.)
> in ways that is most effective for their WG for their current
> work items. While some chairs are certainly better than others
> in doing this, I believe this decentralized approach has a lot
> of value and is generally working OK.

Lars,

While I understand and believe in the value of a decentralized
approach, please be careful about "always" because, in years
past, we operated with a much higher level of AD accountability.
That approach has a lot of value too, particularly because
individual ADs and the IESG are selected by the Nomcom and
accountable to the community while WG Chairs are not.  "Trust
the WG chair", carried to an extreme, means those chairs,
especially the subset of them who have run WGs through one or
more changes in AD, are (at least absent appeals) ultimately
accountable to absolutely no one.

That creates even greater risks with today's IETF than it did a
decade or two ago because, with an increasingly mature Internet
and perhaps inevitably a larger fraction of WGs whose focus is
on enhancing, extending, and/or tuning earlier work rather than
on fundamentals, two things happen: IETF LC becomes less
reliable for detection of issues a WG might not have considered
because even reading an extension or enhancement document may
requires deep understanding of the underlying protocol and the
design decisions that contributed to it [1].  Second, WGs tend
to get more homogeneous in views (or at least general
assumptions) and many WG Chairs with very long tenure and low
external accountability become more focused on getting work
through rather than trying to be sure that all issues and
perspectives (especially ones without active advocates within
the WG) are considered.  The latter makes us more dependent on
IETF LC at precisely the time that the other considerations make
it less reliable.  

And, as evidenced by the number of substantive issues that IESG
members raise after IETF LC successfully completes, _that_ puts
more burden on the IESG (and not just the AD responsible for the
particular WG) to have deep contextual and technical
understanding of the documents being reviewed... or objections
end up being raised based on, for lack of a better term, gut
instinct.

> Involving the ADs in this process might seem attractive in
> terms of oversight and/or to establish a common approach - but
> it also further increases the AD workload (c.f. the current
> discussion on the that). There are severe downsides to that.

Sure.  But I am ultimately suggesting two things.  One is that
there there are tradeoffs between more oversight and
responsibility for common approaches on the one hand and AD
workload on the other.  I hope you and the IESG see things that
way rather than as one or the other being absolute.  The second
is that more oversight might lead to less overlap in WGs and
even fewer WGs that are not really productive in terms of a
better Internet.  In particularly, if the reality is that a
particular change or extension is going to go through the system
with little or no review by non-specialists, then maybe we need
a special review team of those specialists and not a WG for
that.  And, of course, reducing the number of WGs is almost
certainly the most effective way of decreasing AD workload while
retaining the quality of IETF output (or even improving on it).  

Much of that is a special case of "work smarter, not just harder
or more hours".

best,
   john


[1] As a thought experiment, how many reviews have you seen
lately that say, essentially, "I reviewed this document for Area
XYZ, but I don't have a clue about its substance so found the
following editorial or organizational problems"?