Re: Front-end delays
Thank you for a series of reasonable suggestions.
My thoughts are inline.
There has been a fair amount of effort in accelerating the tail end
of the document process, i.e., after IETF last call. It is unclear
whether this has succeeded (as there don't seem to be any published
measurements), but I believe that the main problem with timeliness
is now in the WG process itself. Each case is unique, but there are
a disturbing number of modest-size documents that have taken 3 to 5
years from first individual draft to RFC. In a separate message,
I'll have more process- and people-oriented suggestions, but here
are some simple things that would at least allow us to measure, and
thus manage, delays:
Whether "the main problem with timeliness is now in the WG process
itself" is true or not, it is worth removing systemic sources of delay
in the WG process.
(1) Problem: Currently, the mapping between charter items and drafts
is often vague and those not intimately involved with the process
are often left to guess which draft(s) meant to address which
Proposal: Once the I-Ds have been created, the charter is updated to
reflect I-D names. This doesn't have to be done for every single
draft, but doing such an update once for each IETF meeting seems
easy, as there are no substantive changes.
I like this suggestion. The gotcha is that "I-D creation" is not a
one-step process, as many WGs start out with individual drafts that
get renamed, and even combined, as they are adopted as WG drafts.
If your suggestion is "once the I-Ds have been adopted as WG I-Ds for
a charter milestone ...", I think this would work well - and would
also point out the milestones with approaching dates where the WG has
not even agreed on baseline text.
There is one missing aspect to this - WG charters aren't consistent
between "the WG thinks we're through" (we requested publication) and
"the IETF thinks we're through" (the document has been approved for
publication). We give WGs and ADs a lot of flexibility in specifying
milestones - do we need this flexibility? If charters said, "we have
three dates for each document - WG draft adopted, publication
requested, approved for publication", and tracked these dates, we
would know a lot more about where the delays in the process are than
we know today.
(2) The I-D tracker should keep track of the charter deadline for
each document so that the author and WG chairs can trivially see
whether a document is on track or not. [This item can obviously be
used to automatically generate (1).]
As you say, this is an obviously good thing to do, given (1).
(3) At each WG meeting, it should be expected that after the agenda
bashing, the delivery schedule for each draft is reviewed, maybe
with a red (already out of schedule), yellow (likely to miss
schedule) and green (possible/likely to meet schedule) indication.
Two points here -
The IETF face-to-face meeting is supposed to be our high-bandwidth
exchange opportunity, so reviewing this schedule, with color coding,
can also help us focus on what documents need our attention. This
could also be a time for editors to say, "I need help", or even, "you
need another editor".
A number of WGs are in constant tension between the current charter
deliverables and adopting new work items. If a WG can say, "look,
we're all green, except for two yellow documents that we have a plan
to recover", that's got to help an AD agree to a new charter item.
(4) Just like the RFC editor and IANA provides statistics, each area
should provide an overview of its status at the IESG plenary,
indicating the major work items of each WG and whether the WG is
generally on schedule or behind. Again, one could imagine a color
coding, e.g., if 30% or more of the charter items are "in the red",
the WG is flagged red.
Back in the days of 40 or so APPS area working groups, the APPS ADs
presented their report card for the APPS working groups at an APPS
Open Area meeting, usually on Monday morning. I believe their
classification was something like "on time", "late, but making
progress", or "lost in the weeds". We probably don't need a
quantitative scale for this - simply assigning colors would give the
WGs a heads-up about what their ADs think, and how much "attention"
the WG is likely to get in the next meeting cycle.
Clearly, WG chairs and ADs could "game" the system to make the WG
look on time, by continuously adjusting schedules or by having
deadlines a decade out. That itself, however, would provide an
indication that a WG is not meeting deadlines or is planning to be
really slow. If one wanted to, one could indicate both original and
current deadline for work items.
I can't believe we don't track original planned dates plus current
forecast dates now. We also lose all track of when milestones were
completed, because the dates are replaced by DONE. I would like to see
None of these items make things faster, but they would allow the
IETF management to apply project management tools, such as reducing
scope, replacing authors, or working group chairs, with some
efficiency. More on hypotheses for why things are late in another
Ietf mailing list
Ietf mailing list