OPEN ISSUE: Standards Track

[Ralph Droms]

I agree with Keith about a need for baby steps.  We stumbled
across a sequence of baby steps entirely through serendipity
that seems to have worked well for the DHCPv6 specification (I
say "seems to" because the spec has only recently been
completed so we don't have a lot of history with the results).

The extra baby step we took was to perform two rounds of
interoperability testing between the acceptance of the DHCPv6
spec as a Proposed Standard and the publication of the RFC.
This interoperability testing allowed us to identify a dozen
or so places in the text that could be improved prior to
publication of the RFC: a couple of typos, some text that
required clarification, etc.  We published an I-D (and ran
a WG last call) summarizing the results of that
interoperability testing and the spec has been edited
based on the report in the I-D.

The result will be (I believe; we'll see how the spec fares
in practice) a better Proposed Standard spec.  The analogy
with the current usual path for a spec is, I think, that
draft-ietf-dhc-dhcpv6-28.txt (the I-D that was accepted as
the basis for the Proposed Standard) was roughly equivalent
to the definition (if not the practice) of a "Proposed
Standard".  The version of the spec that will be published
as an RFC is roughly equivalent to a "Draft Standard".  What
was different about the DHCPv6 spec process is that
draft-ietf-dhc-dhcpv6-28.txt had no acceptance and deployment
(which would have happened had it been published as Proposed).

I don't claim that this experience points out a specific
problem (and attendant solution) with the current process.  I
think the experience points out that there may be improvements
we could make to the current process that better achieves
the goal of "experience to prove correctness of a stable
specification" prior to widespread acceptance and deployment.

- Ralph

At 08:38 AM 5/22/2003 -0400, Keith Moore wrote:
>> We need to remove steps, not add them.
>
>actually I also think that we need one or two baby steps, prior to what
>we now think of as proposed, to ensure that certain considerations
>get dealt with early in the design phase, and also to ensure that 
>proposals get wide review long before the Last Call for proposed.