Minutes IETF99: babel
minutes-99-babel-00

BABEL Working Group
Hilton Prague, Prague, Czech Republic.
        Monday, 17 July 2017.
        17:40 - 18:40 Athens/Barcelona Room

Chairs:  Russ White (LinkedIn)
         Donald Eastlake (Huawei)

Minutes: Barbara Stark, Russ White
Jabber: Michael Richardson

Note well was noted.
Chair slides presented:
https://www.ietf.org/proceedings/99/slides/slides-99-babel-agenda-and-status-01.pdf

Agenda
- 4 min. Administrativia (scribes), Agenda Bashing, Chairs
- 4 min. Status, Review of Milestones, Chairs
- 12 min. Mandatory sub-TLVs in Babel, Juliusz Chroboczek
draft-ietf-babel-rfc6126bis
- 10 min. Unicast Hellos, David Schinazi
- 5 min. Information Model, Barbara H. Stark
draft-ietf-babel-information-model
- 10 min. BIER in Babel, Zheng Zhang
draft-zhang-bier-babel-extensions-01
- 10 min. Source-Specific Routing for Babel, Matthieu Boutier
draft-boutier-babel-source-specific-02
- 4 min. Wrap-Up, Chairs

Current draft discussions

Applicability draft was declared to have NOT passed WGLC. It will be
updated and pushed through last call again

-----------

Mandatory sub-TLVs in Babel, Juliusz Chroboczek
Juliusz Chroboczek presented:
https://www.ietf.org/proceedings/99/slides/slides-99-babel-mandatory-sub-tlvs-in-babel-00.pdf

- slide 3: David Schinazi: We have 2 implementations of unicast hellos
  (joking: "1.5 implementations"). Toke and I intend to try to test
  interoperability this week.
- slide 10: Tony Przygienda: You could stick a capabilties TLV that is
  advertised per router -- and those that don't have the capability
  can be ignored/ignore the extensions This is difficult with a
  distance vector protocol, because the capability is not carried
  throughout the entire network (as in link state)
- Slide 11: David: Really dislikes option of bumping protocol
  version. Disagree with need for flag day. You can just do both and
  upgrade routers over time.

On slide 10, comparing the 2 options. Bump the protocol is messy.

Juliusz: It's simpler than you make it. You can always send 2 unless
you set the mandatory bit, then you send 3.

David: But if I send some packets with v2 and some with v3, I think
you end up breaking principals of loop avoidance.

Juliusz: No you can create black holes.

David: Are black holes ok? <general laughter occurred>

Margaret Cullen: Are you confident the implementations we have will
ignore the additional bit?

Juliusz: Yes

Michael Richardson: I'm not thrilled about v2 to v3. What happens when
we go to v4? We still maintain v2?  It seems better to remember what
talks v2 or v3 and send appropriate packets.

Toke Høiland-Jørgensen: I'm not sure it solves the problem when you
have half a network. Not sure the complexity is worth it.

-----------

Unicast Hellos, David Schinazi
Presented slides:
https://www.ietf.org/proceedings/99/slides/slides-99-babel-unicast-hellos-00.pdf

Slide 7: Juliusz: That's not part of the doc and should not be used
for neighbor acquisition.

David: Doc currently says if you only see scheduled hellos
<something>. And this says...

John Dowdell: If you receive multicast and respond unicast, does that
mean everything from thereafter is unicast?

David: In Babel, you keep sending hellos periodically, and you reply
with IHUs. So if you keep seeing hellos and IHUs from your neighbors,
you know they're still there. It's a periodic thing.

Juliusz: 2 comments. One is you are allowed to send "multicast" hellos
over unicast. A hello must be sent to all neighbors. Nothing breaks if
you send to all neighbors over unicast. So new doc version says must
send hellos either to all via unicast, or with multicast. Is it worth
the complexity?  2nd comment is we are contradicting ourselves. Must
parse unicast hello and can drop after parsing it.

Russ: I'd like to take further discussion but not now because I also
have unicast questions.

David: There is ambiguity of sending unicast to all being like
multicast.

-----

Information Model, Barbara H. Stark
Presented slides:
https://www.ietf.org/proceedings/99/slides/slides-99-babel-babel-information-model-00.pdf

Slide 3: Juliusz: In second point, the parameters should be removed.

Barbara: They have been removed, but there was question as to whether
to add them per interface

Juliusz: No

Barbara: Good. Issue can be removed.

--------------

Source-Specific Routing for Babel, Matthieu Boutier
Presented slides:
https://www.ietf.org/proceedings/99/slides/slides-99-babel-source-specific-routing-00.pdf

Toke: I like this.

David: Great work. Support adopting.

Juliusz: Dumpt the dump.

Donald: We will determine if there is consensus for adoption.

--------------

BIER in Babel, Zheng Zhang
Presented slides:
https://www.ietf.org/proceedings/99/slides/slides-99-babel-bier-in-babel-00.pdf

Alia Atlas: Doc should be in BABEL if there is interest. Because BABEL
is still moving and knowledge is broad.

Juliusz: Very interesting work, but not necessarily interesting from
perspective of application. It would be used by data center or
operator. Curious as to how difficult it would be to implement BIER
when you only have the partial view that BABEL gives you.  Next
question is I'm uncomfortable with sub-sub-TLV. Is the needed TLV
widely implemented? Not sure since focus has been on BABEL for
homenet. That makes me wary of adoption.

Denis Ovsienko spoke remotely regarding github implementation.