[tcpm] An update to AccECN draft closing off WGLC

[Bob Briscoe <[e-mail]>]

Michael, and TCPM chairs,

This gives notice of two changes to draft-ietf-tcpm-accurate-ecn-24 
coming up, which will hopefully close off the WGLC:


      1/ ECN++ not the only example

I've just promised Markku (on the tcpm list) a change to the AccECN 
draft that references the text about DupACK detection that we've moved 
to the ECN++ draft:
https://datatracker.ietf.org/doc/html/draft-ietf-tcpm-accurate-ecn-24#section-3.2.2.5.1

* Instead of implying this is only for ECN++, we'll say it's for any 
future RFC that uses ECN-capable ACKs, with ECN++ as just an example 
(AckCC [RFC5690] was another example, but it was an informational 
description of a possible protocol, not an actual protocol spec).

* And, instead of saying ECN++ describes how ECN-capable ACKs ought to 
be conditional on non-ambiguous DupACK tests, we'll say any RFC that 
allows ECN-capable ACKs ought to make it conditional on non-ambiguous 
DupACK tests (with ECN++ as an example).


      2/ Explicit list of conditions inherited from RFC3168

The AccECN draft currently says it builds on the conditions in RFC3168, 
as follows:

     A host in AccECN mode has the rights and obligations concerning the
     use of ECN defined below, which build on those in [RFC3168 
<https://www.rfc-editor.org/info/rfc3168>] as
     updated by [RFC8311 <https://www.rfc-editor.org/info/rfc8311>].

See 
https://datatracker.ietf.org/doc/html/draft-ietf-tcpm-accurate-ecn-24#section-3.1.5

When implementing AccECN, Vidhi Goel has pointed out that one can 
probably guess which RFC3168 conditions still apply (and which don't), 
but different people might guess differently. We think there was also at 
least one condition missing from RFC3168, when a connection falls back 
to non-ECN.

So we agreed it would be best to list the conditions inherited from 
§6.1.1 of RFC3168 explicitly as well as adding the one we think was 
overlooked.

Cheers


Bob

-- 
________________________________________________________________
Bob Briscoehttp://bobbriscoe.net/