Re: [Idr] I-D Action: draft-ietf-idr-large-community-01.txt

[Brian Dickson <[e-mail]>]

Top-reply for clarity.
Apologies if doing so offends anyone.

Here's the rationale for MUST (vs SHOULD):
- there is NO difference in what implementations need to do
- there IS a difference in what operators are able to do if/when collisions occur 

What I believe Tom was trying to articulate is this:

If some sends Large communities which start with HIS_ASN, he needs this MUST to use as a clue-by-four in communicating with the offending party.

If it is only a SHOULD, it removes the only tool available with which to apply pressure. 

(Do the math on the odds of unsigned 32-bit integers above 132K from different number spaces colliding, if those number spaces are not sparsely populated, and assigned in a sequential manner. The odds are "1, eventually".)

Squatting on assigned number resources is a horrible thing to deal with. I personally have had an IPv4 prefix squatted on, and that was something where things were significantly impacted.

It is not likely to happen, and if it does, it is very likely not malicious. However, without this MUST, the usefulness of Large communities are potentially impacted.

Please, let us have the MUST.

It has no impact to implementers, and a huge benefit to operators.

Brian

Sent from my iPhone

> On Oct 11, 2016, at 7:07 AM, Jeffrey Haas <[e-mail]> wrote:
> 
> 
>> On Oct 11, 2016, at 7:09 AM, Nick Hilliard <[e-mail]> wrote:
>> 
>> t.petch wrote:
>>> I want a MUST in there so if anyone does otherwise, we can say No!, not
>>> allowed.
>> 
>> If an operator in the far depths of inner Kaboodlestan uses an IPv4
>> address instead of an ASN, the IETF can disapprove all it wants, and it
>> will make no difference to anyone.
> 
> Somewhat relevant example in other extensions:
> Many implementations treat an add-paths path-id as an integer and you tend to see them showing up as a int in show commands.
> 
> At least one implementation chose to use a BGP identifier/IP address in theirs.
> 
> The spec treats the value opaquely.  Both use cases make sense in their own context.
> 
> End of the day, the protocol cares that it's a 4 byte value.  Users may want a given representational syntax because it makes their life easier in their particular scenario.
> 
> Neither is solely the Right Thing.
> 
> -- Jeff
> 
> _______________________________________________
> Idr mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/idr