Re: [Gen-art] Genart telechat review of draft-ietf-tsvwg-ecn-experimentation-05

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

Just to end the suspense,

On Sep 2, 2017 13:37, "Spencer Dawkins at IETF" <
[e-mail]> wrote:

Hi, Brian,

On Fri, Sep 1, 2017 at 6:41 PM, Brian E Carpenter <
[e-mail]> wrote:

> On 02/09/2017 09:45, Black, David wrote:
> > Brian,
> >
> > Thanks for the prompt review.
> >
> >> Comment: Very clear from the technical standpoint.
> >
> > Thank you!
> >
> >> I understand the desire for brevity, but this bothers me a bit. What is
> >> the reader to make of RFC3168 Section 20.2, for example? My feeling is
> >> that a short Appendix outlining the specific updates would be useful.
> >> There's already too much spaghetti to untangle.
> >
> > RFC 3168 Section 20.2 is the rationale for the ECN Nonce and hence would
> be
> > deleted. Request noted, I'll consult with the draft shepherd and
> responsible
> > AD to figure out whether to do this.
>
> Thanks. It's not intended as a blocking issue.
>
> >
> >> I see no reason why RFC3540 and RFC5622 need to be normative references
> >> (and therefore downrefs). They aren't required reading in order to
> >> understand this draft.
> >
> > OTOH, both are affected by this draft:
> >
> > In reverse order, this draft updates RFC 5622 - that seems to merit a
> > normative reference.    This draft also provides the rationale for the
> > status change of RFC 3540 to Historic, which also seems to merit a
> > normative reference.
>
> Well, my understanding is that a normative reference is needed only
> when the citing document cannot be understood and implemented
> without reading the cited document.
>
> Again it's not a blocking comment - although there is a technical error
> in the Last Call message: it flags the downref to 5622, but not that to
> 3540.
> I don't know if that's a tool error or an AD error ;-).


As Bran is too polite to say, the AD sees the generated Last Call
announcement before it's sent.

I just regenerated the Last Call announcement, and it's still only naming
one draft as a normative downref. But the AD didn't notice that, so blaming
the tools was the wrong answer :-)

It's a 3-day weekend in the US, so any updated Last Call announcement won't
go out until Tuesday, anyway.

So, let's talk about that.

I just re-read the draft, focusing on the places where it refers to RFC
3540.

This draft explains the ECN Nonce experiment inline, although it cites 3540
as its reference, and explains that the experiment required ECT(0) and
ECT(1), but the experiment is over, and hasn't been deployed on the
Internet in any significant way, so we want ECT(1) back, for other
experiments.

It also refers to 3540 when it changes a registration in the Transmission
Control Protocol (TCP) Header Flags registry, but I believe this draft
explains the IANA actions requested well enough that IANA wouldn't need to
read 3540 to carry out the requested action.

I could be convinced that this draft explains why we want ECT(1) back well
enough that it's not necessary to actually read 3540, and since we're
declaring 3540 Historic, we actually kinda hope no one reads it in the
future except archaeologists in cyberspace.

But unless that reference really is informational, it remains a normative
downref that was omitted from the Last Call announcement.

David, I don't mind re-issuing the Last Call announcement if you think
still think the reference is normative, but you might want to take a look
at the text in the draft, and see how badly it needs to be normative,
before letting me know.

Spencer


Mirja and I talked this morning, and the score is

It's normative:2
It's informative:2

between me, Brian, David, and Mirja, so I took that as a sign that moving
that reference to Informative References isn't obviously ok.

I edited the Last Call text and re-issued the Last Call requests for both
the draft and the status change document, so they stay in sync.

The Secretariat was likely off today (US Labor Day), so it might take a day
for the Last Calls to pop out.

Thanks to everyone for your help.

Spencer