Re: [Idr] Éric Vyncke's Abstain on draft-ietf-idr-bgp-ext-com-registry-04: (with COMMENT)

[Jeffrey Haas <[e-mail]>]

Éric,

On Wed, Dec 01, 2021 at 02:00:32AM -0800, Éric Vyncke via Datatracker wrote:
> While I am not aware of the history about the use of those "experimental" code
> points and while I am aware that changing the code points in running code is an
> impossible endeavour, I wonder how those code points were used in production
> networks (my guess) while keeping the status of 'experimental'. I.e., the code
> points 0x81 and 0x82 are noted as 'experimental' but RFC 8955 is standard
> track. I am balloting ABSTAIN as I miss the history.

You have all of the relevant history.

Once a code point is well deployed in a protocol with a scope like BGP, it's
basically a done thing.  Even if there was a decision to try to move to new
code points out of that range to a standards track allocation, there would
be deployed implementations using the existing code points.  A migration
strategy would be required, and would likely not go well.

A related example of where this has failed in prior BGP history are in BGP
capabilities:
https://www.iana.org/assignments/capability-codes/capability-codes.xhtml

See capabilities 128-131.  Many implementations find themselves having to
support the old code points as well as the ones they moved to.

> ***Bottom line for IANA/IESG: what can we learn from this apparent "mess" (I am
> afraid that my limited English vocabulary is unable to find a more positive
> word)?***

The IESG is encouraged to review a presentation I gave several IETFs ago on
the matter.  I'd suggest a form of this be integrated into the general IETF
training material, perhaps through emodir.

https://datatracker.ietf.org/meeting/97/materials/slides-97-idr-code-point-management-02

I'm happy to speak to the IESG at large about the issue.

> I also wonder why "0x80-0x82 | First Come First Served" since this document
> confirms the allocation ? What am I missing here ?

The update is to change the allocation mechanism from experimental to FCFS.
It opens up other allocations within those ranges to general use.

-- Jeff