Minutes IETF113: mpls
minutes-113-mpls-00

MPLS WG Draft Agenda for IETF 113
Meeting : IETF113, Thursday, March 24, 2022

Location: Vienna, Austria (and online), Afternoon session I

Chairs:
Loa Anderson loa@pi.nu,
Tarek Saad tsaad.net@gmail.com,
Nic Leymann N.Leymann@telekom.de

Secretary:
Mach Chen mach.chen@huawei.com

Slides:
https://datatracker.ietf.org/meeting/113/session/mpls/

HedgeDoc for Notes Taking:
https://notes.ietf.org/notes-ietf-113-mpls?both

Meetecho:
http://www.meetecho.com/ietf113/mpls/

Jabber:
xmpp:mpls@jabber.ietf.org?join

1. Chairs Intro (Agenda Bashing, etc.) - Start Time: 13:00
Duration: 5 mins
Presenter: WG Chairs

[Zafar] Not see a draft for the SPL Reuse policy slot, the slides is not
technical, how is this topic selected as a WG agenda item? [Tarek] The DT had
worked on it and Tony had given presentation talk on it, it’s better to let
more people to listen to this rather than just the DT. [Loa] I asked Tony
whether it’s worth to write a draft. [Tony] Would love to write a draft if
people think that is necessary.Otherwise, why bother. [Loa] Regarding SFL
drafts, as sherperd, will take care after the meeting. [Rakesh] WG LC requested
for 6374-sr draft. [Tarek] All requests received are recorderd.

2. MPLS Data Plane Encapsulation for In-situ OAM Data - Start Time: 13:05
ID: https://datatracker.ietf.org/doc/draft-gandhi-mpls-ioam/
Duration: 10 mins
Presenter: Rakesh

[Greg] iOAM encapsulation in MPLS is one of the use cases of MIAD, could you
pls clarify what type of encap this draft proposes? [Rakesh] G-ACh based and
reuse the TTL field. [Greg] Are you proposing multiple solutions? [Rakesh] No,
there are indicator and data, it’s inline with two things that are part of the
MIAD. [Greg] But G-ACh is not part of MIAD or not part of the direction that
the Open DT is working on. [Rakesh] This is based on MPLS experts’ suggestion
to use G-ACh. If MIAD adopts different EH header, then it can be adopted
according the WG process. [Tarek] Will an iOAM header be inserted by the
ingress node or any node? [Rakesh] Be inserted as part of encapsultion by the
encap node. [Tarek] How to handle the case where there is an existing iOAM
header and people try to insert another iOAM header? [Rakesh] Two opitons: 1)
as part of a new encapsulation, or 2)appending iOAM data to the existing iOAM
header. [Tarek] Concern about the Egress LSR tries to remove the whole iOAM
header and you want part of it to persist. Will read the draft. [Rakesh] Good
comment.

3. Label Switched Path (LSP) Ping for Segment Routing (SR) Path Segment
Identifiers (SIDs) with MPLS Data Planes - Start Time: 13:15 ID:
draft-xp-mpls-spring-lsp-ping-path-sid Duration: 10 mins Presenter: Xiao Min

[Tarek] Is the PSID is only relevant at the egress or transit LSR nodes?
[Xiao] Only the egress node.

4. Deprecating the Use of Router Alert in LSP Ping - Start Time: 13:25
ID: draft-kompella-mpls-lspping-norao-00
Duration: 10 mins
Presenter: Kireeti

[Greg] Why use a range of loopback addresses for IPv6 case?
[Kireeti] To add entropy for ecmp purpose, will consider your question and get
back to you. [Greg] Why not just recommand using a single one? [Kireeti] Then
you will loose entropy. [Greg] Soure port number is an option. [Kireeti] True.
[Loa] Which RFC are you tring to update? It’s abvious for updating rfc7506 and
rfc8029. Isn’t rfc6398 is aslo updated? [Kireeti] No, rfc6398 does not say
anything about the paritcular use that we are talking, it says in general
router alert in IP packet is not a good idea. [Loa] That’s not for MPLS?
[Kireeti] No. [Loa] Can you clarify that? [Kireeti] Sure.

5. Policy on MPLS Special Purpose Labels Reuse - Start Time: 13:35
Duration: 10 mins
Presenter: Tony Li

[Rakesh] The conclusions of claim 1 and 4 are false.
[Tony] I disagree with you.
[Darren] Comments about slides 3，new SPL benifits the same way of walking
through the label stack and detecting of indicator, we can disucsss more detail
on the list. [Zafar] Coments about claim 2, Jags’ solution explains how to
reuse EL to reduce label stack, it works very well. If you add a new SPL, you
add two addition labels. My point is that there is no support that claim 2 is
false. [Tony] There are two solutions that carry EL in the MIAD, you don’t need
a separate instance of the entropy label. [Zafar] In some cases, e.g.,
migration to the new SPL, it may result in carrying four labels for a simple
thing. [Tony] I disagree, if you are using MIAD, you have room to put the
entropy in it, and no point in adding an entropy label too. [Zafar] How would
you do migration if there are leagcy devices? [Tony] For MIAD, it always
assumes that there will be signaling protocol to select MIAD capable routers.
[Bruno] Comments about the summary, would you clarify that backward
compatibility applies to draft-decraene? [Tony] Yes. Your draft does not add
LSEs onto the label stack. [Bruno] Would like that summary to be part of your
summary. [Ketan] Would the MIAD be encompassed everything? [Tony] I can’t speak
to what all proposals are going to do. Of the proposals, I have seen that there
are opportunities for adding a lot of various thing onto one SPL. And I am
assuming that we’re going to cram as much functionality into one SPL as we can.
[Ketan] Which of those two, could you refer to that, I see only proposal, the
Jags’ draft that covered entroy, maybe I missed others. [Tony] The FAI aslo has
carried entropy. [Ketan] Do you see any problem of the MIAD proposal. [Tony] I
have no problems with that. [Loa] It’s better to have draft. [Tony] Since it
appears to be controversial, I would be happy to write a draft. [Wim] Most of
the claims are not in the context of bruno. If we change too much, it’s going
to take long time before it’s adopted. We have to probably segment the
extensions to two phases. One is what we do with the minimum effort, and
considering what type of extensions can we support using legacy hardware. What
is the minimum set of functions which can be done with legacy hardware? I’d
like to separate the problem statment space into two catergraies. One for
minimum set that allows us to have something very quickly, the other is for
longer term which is much more extensible. [Tony] It takes some time for MIAD
to get adopted and get out there, there is no question that having all the
functions will take an arbitrarily long amount of time. The only quesiton on
the table is: does reusing ELI save us any time in geting deployed, and I don’t
see that it does anything towards that. [Wim] My point is for a minimum set of
funtionality, I think it does. [Tony] Somebody needs to explain that, because I
don’t see that it matters one way. [Wim] I think bruno’s draft does that.
[Tony] Bruno’s draft only talks about cases where we can just reuse ELI and not
add additinal LSEs onto it. [Tark] Pleas take it to the list. [Tarek] Just a
sharing a concern, if we put ancillary data in the EL, then it will break
hashing. [Kireeti] If we do this hack, even Bruno’s draft is still a hack,
we’re not making forward progress, we really want to make progress in MPLS.

6. MPLS Extension Header: - Start Time: 13:45
draft-song-mpls-extension-header
draft-song-mpls-eh-indicator
draft-andersson-mpls-eh-label-stack-operations
draft-andersson-mpls-eh-architecture
Duration: 15 mins
Presenter: Haoyu Song

Haoyu presented and no time for discsuion.