[OpenAFS-devel] Re: OpenAFS internal protocol standardization

[Derrick Brashear]

On Tue, Mar 15, 2011 at 1:12 PM, Andrew Deason <adeason@sinenomine.net> wro=
te:
> On Tue, 15 Mar 2011 12:05:42 -0400
> Jeffrey Altman <jaltman@secure-endpoints.com> wrote:
>
>> It is true that anyone can submit a wire protocol that is going to
>> become an IETF governed protocol as an information RFC. =A0Simon's point
>> is that AFS3 protocols are not to be governed by the IETF and
>> therefore cannot be submitted using the IETF processes.
>>
>> The Independent Submission Stream process is what AFS3 or OpenAFS
>> protocols require.
>
> Does the Indepentent Submission Stream not fall under the category of an
> "IETF process"? I know it's not the normal process of IETF-handled
> standards with working groups and such, but it's still a process that
> involves the IETF and the IETF specifies some of the procedures for.
>
>> > I could see it not being part of the IETF process if nobody wants to
>> > do that; just the format is something useful. My concern is that
>> > sometimes I get the vibe of "oh god standardization is _so slow_",
>> > so if internal protocol work doesn't require the same timeline and
>> > review and stuff, people are going to try to rush through it. So it
>> > doesn't really matter if it's an IETF process specifically, but
>> > something in that general direction seems helpful.
>>
>> The format of an I-D is fine. =A0Derrick's point is that the place for
>> the document series to live for OpenAFS specific protocols is in the
>> OpenAFS doc tree. =A0Don't publish them to IETF. =A0Just submit them to
>> the doc tree via gerrit.
>
> ...where they can be reviewed by 1 person and then merged in a few days
> when no activity occurs? That's not a slight against the current gerrit
> system, but the perceived norm/average for code changes does not seem
> appropriate for protocol changes. Some separate process or clear
> guidelines would prevent any such parity.

we can't force review. we *should* for drafts set up a special queue
and establish
1) a minimum timeline
2) a list which gets notified when a submission is make

if there's one positive comment, of one, well, technically that is a
consensus. a crappy one, but...


--=20
Derrick