Minutes IETF112: tsvwg
minutes-112-tsvwg-01

TSVWG IETF 112

Monday 8th of November 2021
Notetaker: Magnus Westerlund

WG Status from Chairs:

The Note Well was mentioned.

Document status
With IESG:
draft-ietf-tsvwg-rfc4960-bis Document Shepherd: Gorry
Drafts beyond WGLC:
draft-ietf-tsvwg-ecn-encap-guidelines (Writeup Needed) Document Shepherd: David
draft-ietf-tsvwg-rfc6040update-shim (Writeup Needed) Document Shepherd: David
WG Milestone Review

WBA Liaison Request - (DSCP/5QI mapping) A response was sent in June - David
3GPP Liaison Request - (SCTP & DTLS) A response sent in May - Gorry
GSMA Liaison Request - (Multipath DCCP) This was received in May - Gorry

Chairs: Announcements and Heads-Up
Work related to other WGs:
Maintenance of TRAM WG RFCs would be done by TSVWG.
New proposed work:
draft-piraux-tcpls (See TCPM)
draft-romo-iccrg-ccid5 (Please discuss on ICCRG list)

2.1 Martin Duke: IETF Code of Conduct (5 mins)
Reminder about the Code of Conduct (RFC 7154).
Think about how you can improve yourself.
Bob Briscoe mentioned that he gets as much L4S emails privately
as on the mailing list, many are not wanting to dicsuss on TSVWG mailing list.

Transport WG Drafts: ECN
3.1 Wes Eddy update on outcome of L4S WGLC (10 mins)
- Very strong overall support.
- Several well-articulated concerns.
- Now working through these concerns.
- Some questions will require more work;
- Possible plan to obsolete RFC 3168;
- Question about status related to option 3 of RFC 4774. David Black clarified
that there needs to be clarification, as L4S is not fully accomplishing Option
3; - Confirm WG consensus on acceptable risk; - David Black: Dual-Q is just one
potential AQM mechanism for L4S, and there can be other options. Therefore the
bar for experimantation is lower for this ID. - 5.2 Bob Briscoe: L4S ECN drafts
draft-ietf-tsvwg-ecn-l4s-id draft-ietf-tsvwg-l4s-arch
draft-ietf-tsvwg-aqm-dualq-coupled

Bob Briscoe presenting the updates from the WG last call.
They as editors should have addressed all the issues that they think
can be addressed, so if your issue is not addressed it has been missed;
or may have been taken as a comment on the approach not the spec.
David Black requested that the authors send an email noting which
WG last call feedback they consider to not be addressable.

Stuart Cheshire summarized his comments on L4S on TSVWG mailing list.
Jonathan disputed Stuart’s characterization of RFC 3168 as having failed.
Jahe Holland, commented that there are 10s of million that have 3168 ECN
deployed, although they don’t mark frequently. However, hw agrees that
classic ECN has failed. The core question is what to do with these
existing nodes.

3.3 Greg White: L4S Operational Guidance
draft-white-tsvwg-l4sops

Greg makes a quick summary with a highlight of changes.

If time permits:
Greg White: NQB
draft-ietf-tsvwg-nqb
No time was available.

Session II
Friday 12th November 2021
Notetaker: Jonathan Morton

Agenda Review
Note well
If you aren’t familiar with this by now…

Chairs/Michael Tuexen: Key updates since WGLC of RFC4960.bis
Updates pending from late reviews, progressing with IESG.
e.g. handling tiny rwnds, various other small details.
This ID is on IESG telechat agenda for December, if you wish
to review the latest changes please do so before end of November.

Transport : SCTP

5.1 Michael Tuexen: NAT alternatives for SCTP
draft-ietf-tsvwg-natsupp (returned to WG)
- No text changes this time, but major issues to discuss.
- Two possible fixes: improve current method or always fallback.
- Or try Claudio’s alternative proposal.
- Or drop the spec on SCTP-NAT entirely.
- Magnus: promises to think about it.
- Claudio Porfiri: I am also willing to help.
- Christian Huitema: NAT inherently exposes spoofing attack surface, will
supply scenario in e-mail. Chairs: Please try to see if there can be a combined
ID that might be useful for both use-cases.

5.2 Magnus Westerlund: RFC6083.bis (10 mins)
draft-ietf-tsvwg-dtls-over-sctp-bis
- 2x IPR disclosures from Ericsson.
- Goal: To enable large DTLS-protected messages.
- Requirement of very long-lived SCTP sessions, periodic mutual re-auth.
- Solution: parallel/overlapping DTLS connections.
- New session set up from scratch, meanwhile old session still usable until
messages completed. - Call for feedback, targetting WGLC fairly soon. - IPR
concerns raised, Michael Tuexen states open-source implementations likely
impossible. Chairs: We would like the WG to explicitly consider the IPR
implications on the proposal to obsolete an existing spec., and to publish a
new PS. We expect this discussion at the next IETF meeting, after people have
considered any implications.

Transport : UDP
6.1 Joe Touch: UDP Options
draft-ietf-tsvwg-udp-options
Gorry presenting slides from Joe.
- The OCS and FRAG options were moved to consistent positions in the datagram.
- The UNSAFE option clarified.
- The FRAG option is mandatory, at least 2 fragments & 3000 byte MRSS; higher
limits may be negotiated. This follows discussion at the WG Interim, and Joe
would like to know any WG feedback. - per-fragment and per-segment options
specified for the wire protocol. At least a per-fragment OCS would be needed to
ensure NAT traversal and DPLPMTUD. Per Datagram options treated like any other
datagram. - C. Heard: Padding mechanism needed to extend existing datagrams for
DPLPMTUD probing? Are per-fragment options really required?

6.2 Tom Jones: dplpmtud for UDP Options
draft-ietf-tsvwg-udp-options-dplpmtud
- Will correct text on padding, appending zero bytes should be enough.
- Will clarify other things too.
- Gorry: This spec is intended to progress in parallel with UDP Options.
- David Black: The content has already gone through some review, so should go
quickly.

Transport : DCCP
7.1 Multipath DCCP
draft-ietf-tsvwg-multipath-dccp
- Development on Github.
- Basically a 3GPP aligned spec, work ought to complete in about a year to
align with 3GPP timeframe. - Substantive updates on options, reordering,
fallback, authors, handshake/auth. - Also various editorial updates. -
Handshake for multipath is now 4-way, as 3-way had a race condition. - Would
prefer to optimise back down to 3-way, two ideas proposed for discussion. -
Gorry: I suggest implementing one option and see how it works, maybe this could
be an interim topic? - Tianji Jiang: two-flow versus three or more flows? -
Markus: two-flow examples are just for illustration, protocol generalizes to
more flows.

Differentiated Services
8.1 Greg White: NQB - this was not taken on day 1
draft-ietf-tsvwg-nqb
(Reduced to 5 mins for time)
- Current discussion about NQB requirements, particularly permissible sender
rates. - Normative recommendation to NOT use NQB if send rate over about 1Mbps
or few pkts per RTT. - Greg suggested replacement with a definition that scales
better with time and link capabilities; this would need to be discussed
on-list. - Concerns about variable-rate links and suitability for particular
latency-sensitive applications. - Question (on list): is proposed text an
improvement, and how should it be quantified? - Should both NQB DSCPs have the
same name in IANA? David thought the suggested IANA approach was heading in a
good direction. Show of hands: 6 people said they had read a recent version of
the draft.

8.2 Ana Custura: DSCP Considerations
draft-ietf-tsvwg-dscp-considerations
- Cut for time.

All Other Business
No time.