On Aug 1, 2011, Keith Moore wrote:
Perhaps. But it's difficult to escape the impression that this is another
example of IETF failing to solve an important problem by focusing on a
portion of the problem that's easy to solve, and ruling the difficult part
out of scope for the time being. Repeat as needed; you can always partition
the remaining part of the problem again.
.....
Does it follow, then, that the Right Thing to do is to avoid building any
other parts of the system (even, say, the reputation service query protocol)
until the easiest part is finished?
No extreme is likely to work, because this is essentially an optimization
question for standards processes. We worry too little about the
opportunity cost of the passage of time, so we fight time-consuming battles.
We should instead be trying to build an optimal pipeline of incremental
progress in a generally positive direction, acknowledging inevitable mistakes
along the way.
I'd like to view the IETF as a standards factory, with a certain optimal level
of throughput that we're trying to achieve. As with, say, a sausage factory,
you don't just take in a single opinion, chop it up to bits, and deliver a
snack-sized bit of RFC before moving on to process the next opinion. The
economic benefits of parallelism are so large that they dwarf all but the worst
mistakes, so we accept the inevitability of a flawed process that lets a few
bugs get through. As a technologist I hate that(*), but it's true. And it
means that, if you can look back on a standard you worked on a few years ago
and not have any regrets, you spent too much time on it.
So while we obviously need to focus on smallish, core chunks of technology like
DKIM before devoting a lot of effort to building on top of them, we also need
to start the latter before the former are completely finished.
If anything, we're late. DKIM is almost a decade old. It's pretty good, and
we've long since reached diminishing returns. There's plenty more to do. It's
time to move on. -- Nathaniel
_____
(*) As a vegetarian I just feel smug. :-)
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf