On 17/04/2014 03:32, Melinda Shore wrote:
...
> I'm not sure that the IETF should get more directly
> involved with supporting implementation efforts but I
> do think that we should recognize that work coming
> in that's already been implemented tends to be more
> fully-baked and privilege that in some way. One worries
> about that encouraging crappy implementation, but I'm
> not sure that crappy implementation is much worse than
> no implementation.
I don't think we should privilege it particularly. We have at
least 3 modes of working, each of which has its value:
1. Preemptive standardisation: problem statement first,
then requirements, then design, then standardise, then sit
back and wait for implementation.
2. Parallel standardisation: there's some design work and
maybe some code, then retro-fit requirements, write the specs
as the code evolves, then standardise and hope that the code is
updated to match the spec.
3. Retroactive standardisation: there's some running code that
got itself deployed; write the v1 spec to be compatible with
the code, call it a standard, then develop v1.1 or v2 and
hope the code is updated.
We've done all three, and all with varying degrees of success.
I could give examples, but that would just start arguments ;-).
It isn't reasonable to expect all standards to fit the same
model. And people who don't have the desire to work in any
of these models may well prefer to quit standards work (but
still contribute code to one of the three models).
Brian