Minutes IETF116: pals
minutes-116-pals-03

Meeting Minutes Pseudowire And LDP-enabled Services (pals) WG
Date and time 2023-03-28 00:30
Title Minutes IETF116: pals
State Active
Other versions markdown
Last updated 2023-04-06

minutes-116-pals-03

PALS/MPLS/DetNet Joint Meeting IETF 116

Meeting Information

Tuesday March 28, 2023 - 9:30-11:30 Meeting/Yokohama time
Room: G302
120/120 min allocated
Max participants observed: 71

Chairs: Dave SINICROPE, Nic LEYMANN
Secretary: Dave SINICROPE
Minute-taker: Dave SINICROPE (+attendees on CodiMD/HedgeDoc)

Note: all persons with questions and comments (including those
physically present) must line up into the virtual queue on Meetecho to
be recognized
***

Agenda

1. Chairs Intro (Agenda Bashing, etc.)

Duration: 5 min
Chairs: Dave SINICROPE, Nic LEYMANN (for Stewart BRYANT and Andy MALIS)

Dave went through the intro slides, noting the primary purpose of the
session and in particular the discussion on the future of the open
design team
Note takers - add input into Hedgedoc and will be incorporated with
Chair's notes for proceedings.
Chair noted the session is being recorded.

Non-MNA Session

2. https://datatracker.ietf.org/doc/html/draft-schmutzer-pals-ple

Duration: 10 min
Presenter: Christian SCHMUTZER
Private Line Emulation (PLE) - Give an overview of improvements since
revision -00 in preparation for an upcoming WG adoption call
asking for WG adoption and feedback.

Sasha: Transparent handling of all Layer 2 Control Protocol messages
mentioned in the Synchronous Ethernet use case in the Intro of the draft
may result in incorrect reporting of the recovered clock quality (which
is carried in dedicated L2CP messages). IMHO this should be explicitly
discussed in the (currently missing) Applicability Statement section of
the draft.
This comment should not be considered as blocking WG adoption of the
document.
Christian: we are trying to do something similar to an optical transport
network and the optical network would have ehternet pass transparently.
some of the assumptions for clock quality may not be present. Try to
addressed in text on clock recovery and that requirements must be met by
an implementation. Could add something more about the endpoints and loss
packets, etc. We can discuss and would consider input. The boundaries
have been stated.

MNA Session -

3. Open DT Report

Duration: 10 min
Presenter: Tarek SAAD
5th report from the DT
Tarek went through the slides
Tarek highlighted the working group adoption polls for the MNA header -
wide support (draft-jags) It was noted the post stack data did not have
support and was removed from the document to be dealt with separately.
The document on the extension headers was discussed. There was only
minimal support and the decision was to not adopt the document.
It was noted that the additional Editor (Greg Mirsky) has jumpstarted
draft-ietf-mpls-1stnibble.
It was noted that the DT meeting continuation needs to be discussed. See
the Open Discussion below.

Greg Mirsky: 2pts - without questioning conclusions - on the post stack
data, there were technical concerns with the solution about the
complexity of PSD.
Question about last bullet on alternative solutions to PSD. In the
framework and requirements document it notes optional support of PSD
solutions, but are there documented use cases for MNA that can't be
addressed with ISD? If there are cases that can't be addressed with ISD
THEN (and only then) we should pursue PSD solutions. We should not
pursue PSD solutions without a definitive reason.
Tarek: there were technical points raised with PSD, but the majority of
concerns were with IPR poll issues. For alternatives to PSD, one
alternative is not to have an alternative. If we can show there is no
need then we may not need to pursue it.
Andrew Alston: clarification on IPR objections - the objections were not
around IPR but rather the terms of the IPR. A WG can object to the terms
of the IPR but not the validity of the IPR.
Tarek - right we should make that clarification
Weiqiang Cheng- there are use cases that could use PSD but then we see
if ISD could work for them.
Tarek: we've come to agreement that PSD is not high priority right now
Jie Dong: We have looked into PSD and ISD. ISD is useful for some use
cases and PSD is useful for others. Both options should be moved
forward.
See Tarek's prior comment

4. https://datatracker.ietf.org/doc/draft-ietf-mpls-mna-hdr

Duration: 10 minutes
Presenter: Jaganbabu RAJAMANICKAM
Jags presented the slides for WG draft above
There was an overview of the MNA header and formats
Loa: <bad audio will take question to the list>

5. https://datatracker.ietf.org/doc/draft-jags-mpls-ps-mna-hdr

Duration: 10 minutes
Presenter: Jaganbabu RAJAMANICKAM
update on new draft
Jags presented the slides for draft above (same deck as presentation
above)
The slides went through the details of the several different PSD
encodings.
Welcoming comments and feedback
Requestion WG adoption
Joel Halpern: not concerned about formats - if we are going to do PSD
then formats can be worked. The question however is whether PSD is
needed at all. Any PSD solution adds complexity to the label stack. If
needed OK, but so far there has been no clear demonstration of the
need. What problem are we solving?
Greg Mirsky: thanks for addressing questions raised on the list. Share
concerns on complexity and need for PSD that can't be met with ISD. Is
it really required based on analysis of use cases? If there is a
documented use case that can't be solved with ISD MNA, then and only
then should we look toward PSD MNA.
It should also be noted that 0010 may conflict not with assigned labels
but with SFC header first nibble. Will be examining this in more detal.

Jags: we already have a use case document one use case is iOAM -
Greg: It is not a conclusion of WG that IOAM can't be solved with ISD
Rakesh: why do we need PSD? iOAM has multiple option types preallocated,
direct export, edge to edge, etc. - some can be solved with ISD, others
PSD would help e.g., recording of data on every hop there is time
stamping this would be a good use for PSD.

6. https://datatracker.ietf.org/doc/draft-mb-mpls-ioam-dex

Duration: 10 min
Presenter: Greg MIRSKY
IOAM Direct Export in MPLS Network Actions - Share updates to the draft
resulting from aligning the proposal with the current ISD MNA solution
based on draft-jags.
Greg presented the slides
This was presented to the Open DT and will give overview of solution
Uses RFC 9197 header mapped into ISD MNA.
Asking for feedback and WG adoption.

7. https://datatracker.ietf.org/doc/draft-gandhi-mpls-ioam

Duration: 10 minutes
Presenter: Rakesh GANDHI
New revision of the MPLS IOAM draft. The new revision is based on the
new Post-Stack MNA draft (draft-jags-mpls-ps-mna-hdr).
Rakesh presented the slides.
draft was around pre MNA work and has been adapted to use MNA and
draft-jags PSD proposed formats.
Asking for WG adoption - the PS encoding is not a WG doc so doesn't make
sense

Questions on prior 2 drafts
Jags: <comment inaudible>

8. https://datatracker.ietf.org/doc/html/draft-gandhi-spring-enhanced-srpm

Presenter: Rakesh GANDHI
Duration: 5 minutes
Enhanced SRPM draft, now using MNA encoding
Rakesh presented SPRING draft
asking for comments
Greg Mirsky: 2nd slide - goal is to combine session reflector not being
aware of measurement protocol to ensure accurate one way measurement -
for one way the clocks must be synchronized so for one way measurement
the reflector must put the timestamp.
Also, with SR-MPLS how can you do accurate round trip measurements? If
packet sent back on non-corouted path you end up measuring 2 paths. How
can gurantee return path to be co-routed?
Fang Gao: page 7 - delay T2-T1 if the reverse path is native IP. For RT
what is the type of round trip? How to ensure it remains in the fast
path? (question has been asked on the list)
Rakesh: by default the response is native IP, but with SR could put
reverse path.
Fang Gao: If reverse path native IP - will it take some special
destination?
Will this satisfy clock synchronization, specifically time
synchronization? -
PTP timestamps must be in sync between the measurement points
Rakesh: by default reply comes in IP/UDP. Segment-list can be used to
ensure forward and reverse are co-routed.

10. https://datatracker.ietf.org/doc/html/draft-song-mpls-flag-based-opt

Presenter: Rakesh GHANDI for Haoyu SONG
Duration: 5 min
Flag-based MPLS On Path Telemetry Network Actions
Rakesh presented the slides
Greg - There is an MPLS WG adopted draft for alternate marking

9. https://datatracker.ietf.org/doc/html/draft-liu-lsr-mpls-inspection-msd

Presenter : Yao LIU
Duration: 5 min
Signaling Base MPLS Inspection MSD
Yao presented the slides <audio issues>
no questions or comments

11. https://datatracker.ietf.org/doc/draft-sx-detnet-mpls-queue

Presenter: Xueyan SONG
Duration: 10 min
MPLS Sub-Stack Encapsulation for Deterministic Latency Action. Introduce
a new MNA action used for DetNet Latency carried in MPLS packets and
look for comments and co-operation to work on this solution.
Xueyan presented the slides. It was noted that the draft has been
progressed in Detnet.
asking for feedback

12. Open Discussion

Duration: 30 min
a. Future of the Open MPLS DT
Dave went through the Open DT continuation questions and pros and cons
for both continuing and suspending

Tony Li: Open DT calls have been useful - would like to continue with
some kind of call and the administrivia should conform to need instead
of vice versa.
Joel: disagree with continuing the calls. Important that all interested
parties can make calls. Given the list of things noted that need
resolution, they all must be discussed with the WG, the authors have
gone as far as they can. e.g., The WG must agree on wheither PSD is
needed, not design of bits and formats.
The Open DT doesn't conform with IETF procedures.
Adrian: Agree with Tony AND Joel - agree that the DT has made great
progress and could not have been done with whole WG... what is of
concern is that the DT gives a feeling that it is making decisions vs
recommendations. That it is documenting decisions vs documenting
possibilities and getting design and steerage from the WG. Not easy for
all who are interested to attend those meetings. The record of the DT
progression is not coming through enough to the WG. The DT must design
its role within the WG better, communicate better and be clear about
fulfulling its role.
Andrew: 2nd Adrian's comments - even as AD it is not possible to attend
DT calls. Without constant feedback it creates a problem with keeping
the role of the DT to keeping decision making to the WG.
Loa: <audio issues> Tarek relaying Loa - Loa favors DT to bring
recommendation to the WG so decision in the WG. This is process that we
have been following and engaging the WG when a decision needs to be
made. Here at this meeting we are presenting a decision on PSD. DT may
still have a role to make recommendations.
Zafar: agree with Joel - DT did a great job but we should resume WG
process to get the job done to engage a wider wg community.
Jie - agree with Adrian - easier to discuss with small group in DT
instead of having discussion within whole WG. Could do a periodic report
from DT to the WG email list.
Joel: the fundamental disconnect is the difference between "Are DTs
useful" and "What is the DT role?". If there is a specific problem that
needs to be addressed to come to convergence and agreement is needed,
then that would be a use for a design team. However WG decisions, and
all things listed in the Chairs Intro as work to be done, should be
resolved via email and via the WG process. It may be slower but it
involves the entire community.
Tony: DTs are a good things and we allow DTs to make decisions but DT
decisions must be confirmed by the WG. Perhaps the real issue is to make
this clearer.

b. Open discussion - none

Information for the Session