Re: [Idr] Notification Subcode Squatting (draft-ietf-idr-bgp-open-policy)

[Alvaro Retana <[e-mail]>]

Dear WG:

Happy New Year!!

Thank you for the feedback on this matter.

We haven't heard back explicitly from BIRD or FRR implementers -- the two
reported implementations of draft-ietf-idr-bgp-open-policy.  Still, the
confirmed squatting from cisco gives us enough information to take the next
step.


We should request a new subcode for the Role Mismatch Notification.
Authors: please update the draft to reflect the new request.  I will send a
note to IANA alerting them of the situation.


As for the three values (8, 9, and 10) that cisco is squatting on... Let me
first remind everyone that "ownership" of a codepoint is not obtained by
publishing a draft, even if it is a WG document, nor by having an
implementation, much less one that uses different values from the ones in
the draft (offset by one, like in this case), nor by indicating an intent
to continue work on a document!  rfc7120 is clear in the rules and process
for early allocation.  The IANA registries are the only place where the
assignment of a codepoint is officially reflected; that's it.

If there is interest, the WG can decide whether any work moves forward or
not.  Not only does the idr WG have implementation requirements to progress
work, but draft-ietf-idr-bgp-multisession expired almost 9 years ago.  That
is not to say that the work cannot be resurrected -- it can, but it will
need discussion in the WG to do so.  I leave this decision to the WG, lead
by the Chairs.

Based on that discussion, it may be necessary to deprecate the codepoints
officially.  I'll leave that decision to the WG as well.

Thanks!

Alvaro.


On December 15, 2021 at 3:52:51 PM, Alvaro Retana ([e-mail])
wrote:


Dear idr WG:

draft-ietf-idr-bgp-open-policy is currently in IETF Last Call.

This document specifies a new Role Mismatch Notification (code 2, subcode
8).  This value was initially allocated by IANA in 2018 [1].

Unfortunately, Alexander (cc'd) just pointed out that BIRD uses the same
code and subcode combination for a Notification in the case where an
AFI/SAFI is not supported. :-(  As far as I can tell, no such Notification
is specified anywhere.  However, a quick search points at cisco potentially
implementing the same functionality [2] [3] [4] -- some of these reports go
as far back as 2010.


What now?

Given that there seem to be multiple deployed implementations that use this
code/subcode combination, we should avoid using it even if they are
squatting on it. :-(  Also, BIRD was one of the two reported
implementations for draft-ietf-idr-bgp-open-policy.  So I propose that we
request IANA to assign a different value for the new Notification specified
in this document.

Before moving forward, I want to confirm that the WG agrees with the
proposal.  Please take a look at the questions below.  I would like to hear
from implementers at BIRD and cisco (and any other that I didn't find that
might be in the same situation), and from FRR (the second implementation of
draft-ietf-idr-bgp-open-policy).


Q1: Should a new subcode be requested for the Role Mismatch Notification
    specified in draft-ietf-idr-bgp-open-policy?

[ ]: Yes.
[ ]: No, the squatters should update their implementations.
[ ]: Something else.


Q2: What should we do with subcode 8?

[ ]: It should be deprecated (i.e., no one can use it).
[ ]: Write a specification for an "unsupported AFI/SAFI" Notification.
[ ]: Something else.


I will keep this poll open until Jan/5, 2022.


Thanks!

Alvaro.



[1]
https://www.iana.org/assignments/bgp-parameters/bgp-parameters.xhtml#bgp-parameters-6


[2]
https://community.cisco.com/t5/routing/bgp-3-notification-erro/td-p/1572684


[3] https://www.wireshark.org/lists/wireshark-bugs/201607/msg00164.html

[4] https://bst.cloudapps.cisco.com/bugsearch/bug/CSCsu79206