Re: [Idr] Revised proposed IDR charter

[Jeffrey Haas <[e-mail]>]

On Wed, Feb 03, 2010 at 11:03:32AM -0500, Eric Rosen wrote:
> With regard to the paragraph:
> 
>         The IDR working group also has an oversight role for all extensions
>         made to BGP for other uses that may be developed in other working
>         groups. IDR will review extensions made to BGP in other working
>         groups at least at WG document adoption and during working group
>         last calls. The IDR working group will also provide advice and
>         guidance on BGP to other working groups as requested. In some cases
>         the IDR WG chairs may work with the chairs of other working groups
>         and the IESG to move BGP work into the IDR WG instead of the another
>         WG.
> 
> I'd suggest omitting the first and last sentences.  The first sentence adds
> no content, it just provides something to fight about (what does "oversight"
> mean, exactly?).

I would agree with Eric about omitting the first sentence.  I disagree with his
comment on the last sentence.

> On the issue of whether it is IDR's job to "protect the Internet" by
> blocking other WGs from augmenting the use of BGP (or particular BGP
> sessions), that fight has been fought and lost multiple times in the past
> dozen years.

I believe this is less an issue of "protecting the Internet" and more a matter
of maximizing interoperability.  The day a given BGP protocol based spec
deviates enough from the domain IDR cares about is the day that spec should be
forked and the protocol run on a different TCP port.  At that point, it becomes
that working group's responsibility to provide an interoperability mechanism if
some sort of interoperability is desired.

I bowed very quickly out of the MPLS-TP work but I believe that illustrates
what happens when one group takes another's protocol and decides to do their
own thing.  Let's not have that level of squabble within IETF.

Scalability and session survivability are perhaps another matter.  The new
charter helps address that with the following work item:
- Augment the BGP multiprotocol extensions to support the use of                                                              
  multiple concurrent sessions between a given pair of BGP speakers.  

With the addition of that work item, my two major concerns about IDR's
shepherding role for BGP are satisfied:
- Be interoperable
- If you're going to cram something into BGP that's interoperable but may have
  scaling issues for the wider Internet, give me a way to separate that stuff
  so we can keep global IP connectivity working well.

-- Jeff