[rpd] RPD Digest, Vol 222, Issue 45

[Frank Habicht]

Hi,

inline...

On 7/18/2026 10:49 PM, Nia Petronella wrote:
> Dear PDWG,
> 
> I agree with Asamkele, Gugu, and Tshepo, and I remain opposed to this 
> proposal.
> 
> Responding to objectors one by one does not resolve the common concern 
> they have raised. A collision between AS-SET labels is not a failure of 
> ASN or prefix uniqueness,

It is a failure of AS-SET uniqueness.
Which are not among the famous 3 INR that AfriNIC primarily deals with.
But they are important objects in the IRR. Which AfriNIC operates. And 
thus AfriNIC should set the policies under which it operates the IRR.

> and repeating that characterization does not 
> make it technically correct.

Maybe my above statement is not a repetition, and could be technically 
correct?

> The proposal still has not demonstrated measurable operational harm,

Would it be better if someone would have added this to the proposal:
"" MANRS documented the incident with AS-AMAZON back in 2022, 
https://manrs.org/2022/12/why-network-operators-should-use-hierarchical-as-sets/. 
At the time of writing Google's official source for their AS-SET is RADB 
as documented in their PeeringDB entry 
https://www.peeringdb.com/net/433, but AS-GOOGLE currently exists as an 
empty AS-SET in the RIPE DB 
https://apps.db.ripe.net/db-web-ui/lookup?source=ripe&key=AS-GOOGLE&type=as-set. 
""

Would you then agree that the proposal demonstrated harm?

Just because you didn't see it doesn't mean there was no harm.


> why 
> voluntary hierarchical naming is insufficient,

the author posted on this mailing list the number of collisions that exist.

> or why mandatory registry 
> intervention is the least restrictive solution.

This is where you can propose a less restrictive solution.

> Adoption by other RIRs 
> is relevant, but institutional uniformity is not proof of technical 
> necessity.

Harm that was documented is the proof of technical necessity.

> A policy process should examine objections on their merits, not treat 
> repeated disagreement as something to be individually overcome.

Thanks. Was the AI asked to produce a certain quantity of text, so that 
this blah had to be included?

> The 
> registry should support accurate routing information without converting 
> every preferred convention into compulsory policy.

Now, I have a question.

If "the registry" (AfriNIC) gets this information:

as-set:         AS-GOOGLE
tech-c:         AS156-AFRINIC
admin-c:        AS156-AFRINIC
mnt-by:         google_kenya
source:         AFRINIC # Filtered
descr:          AS-GOOGLE


and another registry (RADB) has this information:

as-set:         AS-GOOGLE
descr:          Google
members:        AS11344
members:        AS13949
members:        AS15169
members:        AS15276
members:        AS19425
members:        AS22577
members:        AS26910
members:        AS36040
members:        AS36384
members:        AS36492
members:        AS36561
members:        AS394725
members:        AS40873
members:        AS41264
members:        AS43515
members:        AS55023
members:        AS6432
members:        AS19527
members:        AS26684
members:        AS395973
members:        AS36039
members:        AS24424
members:        AS396982
members:        AS139070
members:        AS139190
members:        AS394699
members:        AS-GOOGLE-IT
members:        AS-MEEBO
members:        AS-METAWEB-2
members:        AS32381
members:        AS36383
members:        AS36411
members:        AS36520
members:        AS394089
mnt-by:         MAINT-AS15169
changed:        arturolev at google.com 20231030  #16:36:58Z
source:         RADB
last-modified:  2023-11-13T16:19:21Z

then what should the person or the software generating a prefix-filter do?

Awaiting this answer.

It's important. Because this is what this discussion is about. Any 
indication that the opponents have understood this will be welcome.

Frank Habicht