Re: [tsvwg] Wanted: recovering IETF process wonk to explain the "Updates:" header

[Spencer Dawkins at IETF <[e-mail]>]

Hi, Bob,

On Sun, Jul 24, 2016 at 9:18 AM, Bob Briscoe <[e-mail]> wrote:

> Spencer,
>
> I searched for " recovering IETF process wonk" and a link to your
> biography was ranked top {Note 1}.
>
> I'm looking for incontrovertible advice on whether I am correct to place
> various long-standing tunnelling RFCs in the updates header of this draft:
> https://tools.ietf.org/html/draft-briscoe-tsvwg-rfc6040bis


As a RIPW (my own words, FWIW), I don't think incontrovertible advice
exists for this. A variety of IESGs have had a variety of opinions about
Updates over time, and even with every IESG there's likely a range of
opinions. I am told in another e-mail that the RFC Series Advisory Group
was talking about the definition of Updates (and the even more dreaded
Obsoletes) in Berlin last week, so , this is not an exact science.
But, since you ask me, I'm the responsible AD and I have an opinion.

My opinion is that you can say Updates for RFC FOO if you want everyone who
implements RFC FOO to implement your draft, too. But rather than leave you
with my vague opinion, I'd say that if you Tell The Reader what your draft
is updating in RFC FOO and why, it will be a lot easier for reviewers to
figure out whether the Updates header is legit.

Sometimes that's obvious. Maybe you're correcting something that has
already resulted in an Errata being accepted - RFC FOO says "meat" when it
should have said "potato".

Sometimes it's less obvious, especially when (as I believe the case is
here) RFC FOO didn't say anything about "potato" at all, but anyone
implementing RFC FOO needs to know whether to add potatoes.

So, your explanation, that RFC FOO said "cook dinner" and didn't give all
the details an implementer needs, would be very helpful, no matter what the
Updates header turns out to say.

Does that help?

Thanks,

Spencer


> I believe it cannot be claimed that these RFCs did not already specify how
> to process the ECN field, because it is not possible to encapsulate a
> packet with a new IP header without creating a new ECN field.
> Therefore, to my mind, definition of ECN field propagation is an update
> not an optional addition to these RFCs (as I said in response to Mirja in
> tsvwg on Friday).
>
> To be clear about the implications of this decision, if the IESG
> eventually approves rfc6040bis as an RFC, if it states that it updates
> L2TPv3 (for instance), I believe it would mean that an implementation of
> L2TPv3 could no longer claim to comply with the L2TPv3 spec [RFC3931] until
> it had implemented ECN propagation compliant with RFC6040.
>
> Cheers
>
>
> Bob
>
> PS. Thanks for the advice to state explicitly how the text of each RFC
> could be updated.
> Will do.
>
> {Note 1}: I was initially worried that an RFC I co-authored was ranked
> second, but I was relieved to discover that we had used all those words in
> a purely technical context: the RFC talked about packet *process*ing and
> *recovering* keys and C.K. *Wong* was a cited author.
>
> --
> ________________________________________________________________
> Bob Briscoe                               http://bobbriscoe.net/
>
>