1. When we start an effort, we do not press for demonstrated
community
need -- but more importantly, demonstrated community interest in /using/
the output. So the folk who work on a topic tend to have no sense of
urgency. (Even when there is a claimed sense of urgency, such as for
STIR,
the work often is not pursued in a fashion that matches the claim, with an
eye towards rapid development and deployment.)
This is certainly true...
But I think there is a second reason in this neighborhood that also relates
to this one --
2. The folk making IETF approvals feel an unfortunate fear of
letting
flawed specifications through the process, even though the fear does not
produce obviously superior results. So we impose high barriers to entry
and
high barriers to completion.
We've lost the art of base spec -- leave other stuff to later. Maybe I'm
just being nostalgic, but I seem to remember a time when we would pass
through a base protocol with extensibility, and then start talking about
extensions on a case by case basis. Now we seem to see 15-20 drafts proposed
in a few months, all with interlocking bits and pieces, totaling hundreds of
pages of text, and sounding more like a bill being presented before some
legislative body rather than a technical specification. These large scale
"boil the ocean" efforts constructed (apparently) by off line meetings
outside the mailing list and the "normal process," are challenging (to say
the least) to even read, must less to fully participate in. When someone
does try to discuss one of these monstrosities on list, the reply is either
"you're stupid," or "you didn't really read all the drafts," or some such,
shutting the discussion down. Of course no-one has really read the drafts --
they're essentially unreadable, and they describe a system of massive
complexity that few people probably understand -- even the authors. I'm
certain each author understand some small bit, but the overall system is far
too complex to be understood by anyone who doesn't have time to dedicate
themselves full time for several weeks to understand it.
This doesn't "improve the speed," as some folks claim--biting off smaller
chunks would actually be faster, as it would increase community
participation, and help us to drive simpler specifications that people
outside the IETF could actually read and understand. If we could get out of
the habit if "boiling the ocean," then we could, possibly, get back to doing
simple things quickly in a serial fashion with a lot of participation. What
we seem to be doing, instead, is a lot of large scale systems that often
overlap in parallel with specifications so complex and so poorly written
that we don't have high overlapping participation rates, which means speed
and innovation suffer, and quality creeps towards atrocious.
At least that's my view of the process at this point.
What we have on the speed front is a culture issue as much as anything else.
:-)
Russ