[MBONED] MBONED minutes from Atl

[Leonard Giuliano <[e-mail]>]

Thanks to Dino for taking notes.  Here are the minutes from today's 
meeting.  Please take a look while things are fresh on your mind and speak 
up if you see anything that should be added or changed.


****


IETF Atlanta MBONED WG - Monday November 5, 2012, scribe: [e-mail]

Status of WG Docs
-----------------
- AMT has taken a lot of time but mostly done
- Multtrans work, 2 drafts adpoted:
   draft-ietf-multtrans-addr-acquistion-00
   draft-ietf-mmboned-v4v6-mcast-ps-00
   Overall, little to no actiivty for Multrans since Vancouver.
- draft-ietf-mboned-mcaddrdoc-04 is now RFC 6676.
- multicat-cdni, authors could not make WG due to Hurricane Sandy=20

Multtrans
---------
Comment from Axel (from DT):
- DT needs 4-to-6-to-4, which means a tunneling solutionc an be used =
versus=20
   a translation solution.

Onjit (sp?) China Telecom:
- Need IPv4 at customer to IPv6 at the service provider.

Lenny: Should we keep translation services in WG charter.
- Toerless, restrict to the solutions that unicast has for transition. =
So
   don't listen to multicast only requirements.
- Stig, don't have to work on them, but just have a need for them.
- Tom Taylor: consider the Map solution. Address+port solution.
- Ron: agree with Toerless, but ... look at uincast transition =
mechanisms
   and see how they break multicast solutions.
- Dino: listen to providers that are deploying v6 cores and will be =
doing
   unicast and multicast service at the same time.
- Toerless: agrees with Dino but look at all solutions and throw out
   what is not needed.
- Raise of hands vote: 75% says keep it.

Mrace Version 2: Hitoshi and WeeSan
-----------------------------------

- Reworded the draft per AD's request.
- Added example how mtrace2 works.
- Change authorship keeping Hitoshi as main author.
- mtrace2's client's location - queries can come from non-adjacent =
clients.

Security concern:
- Stig: should default behavior to not have last-hop routers accept =
query?
- Hitoshi: this was the behavior in last version of design, but change
   for this one.=20
- Stig: mtrace1 is liberal, but should be configurable.
- Gory: let people do what they want to do.
- Toerless: prefer last-hop to forward upstream when it doesn't know =
where
   it comes in.
- Dino: make it plug-and-play, the queries don't create any state so it
   is as harmless as a ping.

- Added Extended Query to allow multiple paths for mtrace to take.
- New types can be extended and a transitive bit has been added.
- Similar draft draft-hout-dp-mtarce-00. May work together.
- Should this be WGLC? Take to list. Not enough reviewers of the draft.
   Sounds like no objection. But taking to the list.

Multicast in the Data Cneter Overview: Mike McBride and Helen Lui
-----------------------------------------------------------------

- Motivation was from the ARMD WG.
- Multicast vital in the data center. And others said wouldn't consider
   multicast at all in the data center.
- Draft gives no recommendations, just says where it stands in the data =
center.
- Multicast apps: Windows Media Servers,  Publish and Subscribe,  Market =
Data,=20
   IP Broker, VRRP,  Overlay, Compute System Monitoring (Ganglia), =
Windows
   Server Cluster Node, Others?
- Dino: add an item for "extending subnets".
- Request the draft to be a WG document.
- 3 people read either first or second drafts.
- 5 yays, 0 nays to accept WG draft. Take to list for final vote.

AMT Multicast Production Implementation: Scott Brown (Octoshape)
----------------------------------------------------------------=20

- Streaming 2mbps to 2 million receivers. Hard to do with CDNs.=20
- Even more difficult outside of US.
- TV sized audiences are coming, high quality to 2-5mbps.
- IPTV deployments have provisioned for QOS Multicast. Trying to bring
   to multiple screens, now how to you bring to 3G/4G that is not =
provisioned
   like the traditional wired network.
- Multicast Design Goals:
   Provisioning and configuring each hop is a non-starter.
   Need an over-the-top solution to allow a simpler way.
   Don't require peering relationships. Too complicated.
   Needs to work over best-effort networks.
   Support adaptive bitrate technologies.
   Don't require telco/operator to install proprietary software in their=20=

   network.
- Major League Gaming (MLG):
   Octoshape and AT&T worked together, March 2011 success event.
   April 2012 went production.
- AMT relays are public so anyone can use them. Events:
   CNN.live (US), NBA League Pass (US), Eurivision Song Contest (EMEA),=20=

   Digturi Ieft Fooball (Turkey), TV Erocr Olympics (Brazil), IPC/ATOS=20
   Paraolympics (London).
- What is Octoshape .... (IPR on this so no notes)
- AT&T and Juniper AMT Relay used.
- Toerless: asks question, phone rings, he leaves room.  ;-)
- AMT Discovery Technique, uses reverse DNS to find sources and AMT =
Relay,=20
   uses internal mechanism to find multicast group.
- Lenny: did you have to write one implementation and ported multiple =
times?
   Answer is yes.
- Scott demoed a CCN app on Firefox, with a Octoshape plugin on MacOS. =
The
   AMT Relay is the AT&T Relay implementation.