IETF 105 Montreal
MBONED Agenda
Thurs, July 25, 2019
3:50-5:20PM
Notre Dame

Jabber Log: https://www.ietf.org/jabber/logs/mboned/2019-07-25.html
Audio log: https://ietf.org/audio/ietf105/ietf105-notredame-20190725-1550.mp3
Video log: https://www.youtube.com/watch?v=-zY8eGHF-k4

Note taker: Jake Holland and Kyle Rose

-----

Notes take in etherpad at https://etherpad.ietf.org/p/notes-ietf-105-mboned

Text (from 8/7/19) pasted below for reference:

MBONED
IETF 105 Montreal
7/25/2019

Minute takers:
Kyle Rose
Jake Holland

3:54 started, note well noted
3:55 agenda

Status of drafts
- driad wglc completed, shepherding in progress
- mcast-problems, minor revisions
- dc-deploy in wglc, please feedback
- deprecate-interdomain in wglc, please feedback
- multicast-yang-model

    - stig: request yang doctor review, before wglc is preferred

    - jake: operators apparently running it, if not obviously broken we should ship it.

- ambi

    - jake: ambi will be coming up again


Stig mtrace v2:
    - problem with conflict in deployed traceroute code.  mtracev2 is the first iana assignment from this space, and hit a conflict.
    - this breaks traceroute
    - toerless: who maintains this?
    - kros: if it's old enough, it's grandfathered in
    - toerless: 
    - lenny: windows doesn't do this, why does it break? (4:05)
    - warren: there's multiple ports used in unix to race, also in juniper
    - stig: need to update and/or obsolete the rfc
    - warren: homenet hit something like this too.  should file errata and also make a new rfc.
    - jake: do we know how to avoid colliding with the chosen replacement port?
    - plan: file errata, reach out to iana, resubmit new rfc, and give warren a heads up reminder.

        - hitoshi volunteered by lenny, did not object.


Charlie Perkins update on ieee802-problems (16:13):
- lenny: doing one more wglc, planning to do "silence == consent" given that it already completed a wglc
- charlie: will send a diff of all the changes from draft -05 to draft -07 for wg review

Kyle (16:21) alta:
- toerless, stig: clarifying questions about tesla

    - kyle: real asymmetric crypto is honestly better if you can have it.

- toerless: why not gcm? (https://en.wikipedia.org/wiki/Galois/Counter_Mode)

    - jake: gcm is symmetric.

- toerless: this is not the right wg, we are not crypto experts

    - jake: secdispatch said demonstrate interest

    - kyle: chairs, what do you think?

    - chairs: AD, what do you think?

    - warren: better to do much of this work where there's security expertise

    - greg: but we can add value by explaining the use case


Jake (16:44) hackathon report:
- stig: possible to use shared libraries/LD_PRELOAD, to use this with existing codebases?

    - jake: socket api important choice to some of this. Not really cross-platform: different sockopts, etc.

    - jake: hopefully TAPS API will be easier to work with than sockets directly. One lesson from TAPS is that three different groups tried to implement a sockets replacement and ended up deciding this is complicated, and have been hammering at it for a few years.

    - stig to point jake at code that does this.


Glenn & Leslie (16:54) mops bof update:
- glenn: smpte (https://www.smpte.org/) uses ietf protocols
- toerless: broadcast equipment has been doing this for a while, and they always think they know better than we

    - leslie: they get confused, don't know where to go

    - toerless: disagree

    - glenn: they seem to think they are confused.

    - by the way, time is a huge topic for them.

    - femi komolafe: agree presenters, customers get scared even of an ip address, but they need answers.

- toerless: scope does seem broken

    - leslie: willing to work on that

    - warren: maybe dispatch/secdispatch is a precedent for something similar

- leslie: trying to help, would be pointing some to mboned of course

    - purposeful cross-pollination

- greg: seems completely complementary
- lenny: look forward to collaborating

Greg (17:10) Bier update:
- time to start paying attention from ops perspective, not just protocol design anymore
- use cases would be interesting
- ops inputs would be valued

toerless: traceroute api is part of the traceroute problem.  the port comes out per packet