[ppml] Proposed Policy: 4-Byte AS Number Policy Proposal

[Geoff Huston]

At 01:34 AM 16/12/2005, william(at)elan.net wrote:

>On Thu, 15 Dec 2005, Bill Darte wrote:
>
> > the leverage to move.  If you think the dates cannot be met,
>
>No, I don't think these dates will be met. Nor does IETF have a good
>record in meeting proposed dates for introduction of new technologies
>or meeting its own deadlines in WGs for that matter...


When you say "the dates cannot be met" don't forget that the industry is 
positioning itself against a very hard barrier - namely the exhaustion of 
the 6 bit AS number space, and the transition into a larger 4 byte pool is 
one that cannot be deferred for anyone's particular convenience. So if this 
date will not be met, the only alternative is ... ? This is not a 
choice-rich environment here, and exhaustion of a finite resource pool 
tends to drive a set of imperatives that, ultimately, are not easy to argue 
with.

The essential element of this proposal is to commence a transition well in 
advance of an exhaustion crisis, and provide some clear signals to vendors 
as to the dates by which 4 byte AS number capability is required in BGP 
products, and clear signals to network operators and admins that if they 
need AS numbers, the dates by which they will need to be fielding these 
4-byte AS number products. Of course if you don't plan to be needing 
further AS numbers sometime after 2009, then this is of no particular 
consequence for you,

> > then those stakes can be driven further out perhaps, but I see the
> > timeline as being a principle quality of the proposal.
>
>I believe its good to mention that specific timeline will be posted
>as agreed by the community. I just don't think specific timeline
>should be set in stone at this stage as part of policy text.

Pragmatically it is not possible to make this process significantly faster, 
and if you are advocating here to make it much slower then all that will be 
achieved is  creating policy stands the strong risk of unglueing itself 
from any form of reality, simply because there is also this issue of the 
timeline of AS number exhaustion to consider, which is still looking like 
happening in October 2010, absent any changes in AS allocation behaviours. 
The entire concept of this policy is to avoid an expensive and needless 
industry crisis through decent advance planning, and, yes, perhaps 
surprisingly, planning typically involves the specification of milestones 
and dates. Through this planning process we can make this transition 
painless and relatively simple. Or we can, collectively, simply stuff it 
up! :-)

regards,

     Geoff