Re: [Idr] Review of draft-ietf-large-community-06.txt

[Geoff Huston <[e-mail]>]

> On 4 Nov. 2016, at 9:34 pm, Jeffrey Haas <[e-mail]> wrote:
> 
> 
>> On Nov 4, 2016, at 4:59 AM, Geoff Huston <[e-mail]> wrote:
>> 
>> I just noted that RFC1997 and RFC4360 had these constraints.
>> 
>> It seems strange to me that an implementation would handle aggregation differently, treating communities and extended communities one way and large communities in a subtly different manner.
>> 
>> Frankly I would prefer to see a consistent treatment of communities in the case of aggregation, and reproducxing the RFC4360 text kinda makes that clear (at least to me)
>> 
>> Omitting it invites different handling and that would be not good
> 
> The relevant point from the thread is that the atomic-aggregate attribute is largely protocol noise.  It's a vestigial organ from the BGP-3 to BGP-4 transition, and a poorly specified one at that.  We shouldn't include its use in specifications, particularly where discussing aggregation.
> 
> The only place its discussion would be relevant is as part of *de-*aggregation of a prefix.
> 

I have to disagree with this position. I am more than happy if you want to push out a spec that updates RFC1997 and RFC4360 and drops the atomic-aggregation conditions and applies the same considerations to this document as well. But having current specs that say “do <this> with communities and extended communities and do <that> with large communities” is bonkers! We should be consistent here and not just deliberately generate inconsistencies here - that is not doing anyone a favour.