[Bier] BIER WG, IETF 98 Minutes

[Greg Shepherd <[e-mail]>]

Here are the minutes from our WG meeting this week. Please read and send
any corrections to me. I'll be posting shortly.

Thanks,
Greg (chairs)

-----minutes-----

BIER WG Minutes

IETF 98 Chicago

March 27, 2017 1300-1500


draft-pfister-bier-mld-03, presented by Stig

Stig:

Early next week we will Take to the list for interest.

No questions from the WG


draft-hu-bier-oam-yang, presented by Fangwei Hu

Stig

We have multicast yang design team.  You should coordinate with them

Greg S:

We need to roll Bier Yang team into that fold

Greg Mirsky:

We should look at connectionless vs connected models to see which one makes
more sense

Greg S.

Are you asking for WG adoption?

Fangwei:

Not now.  Details still need to be flushed out



Echo Request and Echo Reply for Overlay

OAM Header for Overlay Networks, presented by Greg Mirsky

Tony P:

How to deal with MTU that OAM header addition will cause?

Greg Mirsky:

Proactive MTU discovery is important.  Some scenarios are being discussed.
Good question

Greg Mirsky:

Asking for WG adoption in NVO3 group.  Not here.

Greg S

We need input from Bier WG



Early code-point allocation, presented by Tony P.

Eric Rosen:  Juniper

Need code-point for Bier Tunnel Types.  Already requested on list.

Alia:

Is it specified what is needed for IANA.  How solid are the drafts?

Tony:

not squishy

Alia:

If you want codepoints now, watch for the 1-year cliff that you’re
triggering

Eric:

If you already used a value, good time to speak upon these values.

Tony:

Some code-points are being squatted on.

Ice:

Thanks for raising this, we’re checking on SR codepoints & will come back
with BIER values.



Open discussion on WG next steps:

Greg:

We are currently Experimental status.  Deployment experience is a
requirement in charter.  We should ask just for Lab experience. Accept by
raise of hands?   1 person said it was inadequate

Andrew D:

Still wants experimental.  Just playing in the labs.  May never see light
of day.  We have good momentum, though.  Step forward but not sufficient to
remove .  Need production experience.

Greg:

Not all labs are the same. Some labs may be full replication of a
production network

Andrew:

Unless willing to put money-generating traffic, then experimental.  Just a
question of experimental vs not.

Greg:

Only need to be “willing”?  Don’t need to actually do it?

Andrew:

Hypothetical.  Not technical discussion.  Deployment experience is not a
lab.

Greg:

Sense of room.  Worth of effort for standards: Experimental 4. Standards
has 15ish

Alia:

Re-chartering WG needed for Standards track.  Alia needs to take charter to
IESG for approval.  Needs a solid reason.  Process available to take an
individual document to make a Standard.  Even same RFC, but could just give
a new RFC #.

Greg:

Enough interest to work on this document?

Alia:

I would welcome progress on this item.   Did not envision lab experience
would be thought as deployment experience.

Ice:

Input from operators would be good.  Does experimental matter to them?

Lenny:

Creates too low of a bar to call lab as deployment.  Not substitute for
real deployment.

Eric:

What difference does experimental make.

Martin H., DT:

This question doesn’t matter too much.  Have the code and the document
(RFC) is a plus.  RFC is better than draft.  Lab is similar to deployment.

Greg M

What about interoperability.  Testing done in lab.  So should be sufficient.

Luay, Verizon:

95% of what is deployed should be standard.  If standard, then more likely
to deploy..  Depends if you have other options.  We’re a Big Service
provider, though.

Robert R, Bloomberg:

Any change in implementation if code-points change when moved to Standard? For
us it’s important to have an RFC. Experimental or not is not that important.

Alia:

IANA code-points don’t change.  Reason why experimental,  narrow point in
hourglass, if it provides concrete benefits, such that strong motivation to
deploy.  But if it doesn’t have business, driver doesn’t want IETF
standards to pay the price for many years.

Greg:

Can take this forward and can change to standards track in future. (Started
rambling about the history of multicast, why IPMulticast was an
architectural mistake, and why Bier is superior.)

Who wants to assist with document?  (3 hands) I will take to the list to
look for interest in document.

Lenny:

You were concerned of chicken-and-egg with experiments.  However, if it’s
useful it will get used.  So, don’t worry about that.



Adjourn