Re: [Idr] WGLC on draft-ietf-idr-as-private-reservation-00

[Jared Mauch <[e-mail]>]

On Nov 29, 2012, at 1:21 PM, Geoff Huston wrote:

> On 30/11/2012, at 3:49 AM, Warren Kumari <[e-mail]> wrote:
> 
>> 
>> On Nov 28, 2012, at 10:07 PM, Geoff Huston <[e-mail]> wrote:
>> 
>>> 
>>> On 29/11/2012, at 8:26 AM, John Scudder <[e-mail]> wrote:
>>> 
>>>> Folks,
>>>> 
>>>> We have received a request for a working group last call on draft-ietf-idr-as-private-reservation-00. A URL for the draft is http://tools.ietf.org/html/draft-ietf-idr-as-private-reservation-00
>>>> 
>>>> Please send comments to the list by December 14. [*]
>>>> 
>>> 
>>> I am opposed to this draft - increasing the size of a unordered non-unique identifier space is counter-productive.
>> 
>> … and I support it -- some folk have a need for more than 1023 "private" ASNs and *really really* don't want to do the "just reuse the same one and then use something like the 'allows-in' and similar hacks".
>> 
>> Yes, going to an RIR and requesting a few thousand global ASes is possible, but to me seems inelegant.
> 
> ???

I think the issue he raises here is that to provide uniqueness, one could ask for real ASNs, but folks may not or be unable to do this.  If they are not used on the big-I Internet, why should they come from that pool.

> 
>> 
>> Close to 100,000 reserved ones feels overly large to me, but it does line up nicely on a (decimal) boundary.
> 
> But if there is a single need for 100,000 code points in single coherent space then you have to ask how one could ensure uniqueness within the deployment space, as you are now talking about a large number of entities and a large coder point space, and according to your response, the reason why there is a driving need  duplicate an existing uniqueness framework is because the current framework we have for ensuring uniqueness of AS number code points is ... "inelegant"?

I raised this issue privately with someone earlier today as well.  Collisions are likely to happen as someone will either
take the min(private_as_space) and increment, or take max() and decrement.  It's impractical to take a random number from that space, and even if so, does not prevent a collision.  If you have a need for a unique ASN to avoid this case, one should apply for one, even if used on a private network.  In arin region it seems it would be $500 one time + $100/year.  Likely around the cost of those 10G optics or the O&M on the equipment being used so easily justified.

> Obviously I'm unconvinced by such as assertion, and remain opposed to the further progress of this document.

I share this reservation and respectfully disagree with my friend Warren on this topic.

- Jared