Minutes IETF115: grow
minutes-115-grow-00

IETF 115 GROW meeting agenda

Administrativa
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

- assign jabber scribe (job)
- assign minute taker (jared mauch)

Agenda
=3D=3D=3D=3D=3D=3D
ReHash Agenda                                   -  5 min

  * Update on Active WG documents - Job Snijders (10 min)
  * Update on BMP/TLV && ebit - Paolo Lucente (20 min)
    draft-ietf-bmp-tlv && draft-ietf-bmp-tlv-ebit
  * BMP high availability data-collection - Zhuoyaon Lin (15 min)
  * draft-cptb-grow-bmp-yang-04.txt - Camilo Cardona (10 min)
  * draft-cppy-grow-bmp-path-marking-tlv - Camilo Cardona (10 min)
  * BGP Community JSON schema - Martin Pels (5 min)


draft-ietf-grow-bmp-tlv-07 -> 10

draft-ietf-grow-bmp-tlv-ebit-01=20
* adopted as WG doc

draft-lucente-grow-bmp-rel
* add event driven message type to BMP
* complimentary to existing messages

Jeff Haas: You have reached a moment of popularity when everyone wants to p=
ush things into it.  You have a mechanisim where it makes sense.  You want =
to be careful if we want others to attach to it.

The more you add in things like MP extensions, you will have more informati=
on and overloads

Jared Mauch: for the alerting does this live on the BGP speaker?

Paolo: Yes

Jared Mauch: This seems dangerous to have much of the telemetry and alertin=
g on the router=20

Ahmed Elhassany: Change the name of the draft to make it clear there's a ne=
w version coming

Paolo: I will think about that

Presenter: Zhuoyao Lin (Joya Lin)

[shows slides]

Jeff Haas: What timestamp are you collecting for the trigger event, from th=
e protocol or the timestamp from the collector

Joya: the timestamp of the start time of the collector

Jeff: there is no guarantee you will receive the same messages due to order=
ing of messages due to collectors not being in the same state of syncroniza=
tion

Joya: Since it's configured from router -> collector shouldn't they be the =
same?

Jeff: They will eventually (converged example) be consistent, but the trans=
ient state may be different.  Incremental consistency may not be there.

Joya: We could just implement active/standby feature

Chris Morrow: This looks like cool work, it wasn't clear in the slides that=
 this is in-development.  Since BMP data isn't ordered the same way, you ma=
y only be intersted in the ordered stream out of the system.

Severin Dellsperger: Have you thought about using an intermediate service t=
o save the data, or something like memcached?

Joya: I've not heard about memcached

Severin: maybe you use an intermediate service vs redis

Job: can you send a link to Joya for memcached?

Presenter: Camilo Cardona (NTT)

[shows slides]

draft-cptb-grow-bmp-yang: Ideally we would see call for adoption, chairs wi=
ll follow-up on this

draft-cppy-grow-bmp-path-marking-tlv-07: records if path was installed or f=
iltered, optional field for why (eg: local preference) - depends on draft-i=
etf-grow-bmp-tlv and draft-ietf-grow-bmp-tlv-ebit

Presenter: Martin Pels=20

BGP Community JSON Schema

Not a draft, an idea I wanted to put in front of the group

Use-cases:
    Publish at PeeringDB, RPKI(?)
    Parsing by looking glasses
   =20
Q&A:

Jared: we already have https://www.iana.org/assignments/bgp-well-known-comm=
unities/bgp-well-known-communities.xml and similar formats and encoding, do=
 we want to mirror that, or how to discover them etc.  We shoudl be careful=
 that we haven't wanted to standardize types of communities.

Camilo: should we describe this in Yang then convert to JSON schema?

Jeff Haas: I've had a flavor of this discussion with several people, this i=
s a good idea, happy to directly collaborate on this.  May be easy to do in=
 a yang schema.  Jared gets the point that we don't want to be too normativ=
e in how we use communities?  Would suggest something not JSON

Ammed Elhassany: would be nice to look at this use-case in general.

Rudiger Volk: Interesting, when we implemented large communities i created =
something internal called community registry.  I think almost all points th=
at Jared mentioned were covered.  I was told the complexity of the exiting =
system makes it hard.  The notation looks similar to domain name, where you=
 can put in symbols and variables into the string.  You can even generate t=
he regular expressions used in vendor policy languages for matching simple =
communities, or sets of communities.  As far as I can tell, this is working=
, the only thing that hasn't been mentioned right now is we have not addres=
sed how to make a system for global access to definitions.  I'll try to fin=
d some easy examples from the old documentation and share that for people t=
hat are interested, please e-mail me.

Chris Morrow: The communities about the larger BGP message - we ingested al=
l the route-views.org data into bigquery to make it available to researcher=
s.

Jeff Haas: as a follow up, i don't think it's in scope for this, but it wou=
ld be nice if there were consistent naming for the fields.  This is a relat=
ed piece to process this metadata like AS numbers or communities, like VPN =
contexts etc.  having a generic format to decode is good, but international=
ization is also a concern.

Job: Seems there's interested in this space, so we should write a draft.

Drafts in-flight: We have one ongoing call for the WG adoption for the well=
-known community for anycast.  Please comment on-list if you are interested=
.