[arin-ppml] uncoordinated market for IPv4 addresses will cause routing failure

[John Schnizlein]

Changed the subject because this applies to any transfer policy.

I have been looking for justification of a managed market specifically  
based on the need to preserve (somewhat) the hierarchy of addresses  
within the global route table.  If arbitrary transfers would  
significantly scramble the routing tables, that would seem to justify  
undertaking the challenge of operating a market maker.  In such an  
approach, asks and bids would be cleared so that the particular  
address block allocated would better fit the routing hierarchy than  
random (uncoordinated) allocations.  Setting up such a market maker is  
not a trivial undertaking, and would concentrate questions about  
fairness on the market operator, rather than distributing them among  
the parties involved in transfers.

What I have heard is that the fragmentation of global routing due to  
traffic engineering and multihoming has already scrambled things  
enough that random transfers is not likely to make much difference.   
It has been pointed out that transfers are likely to only gradually  
add to the complexity of the global route table, not cause a sharp  
spike.

In the context of similar discussion in RIPE, it seems that an impact  
on the global route table is not expected.
http://www.ripe.net/ripe/policies/proposals/2007-08.html
  "the RIPE NCC does not anticipate that any significant impact will  
be caused if this proposal is implemented"
Is there enough difference between ARIN addresses and RIPE addresses  
to think the conclusion for ARIN would be different?

Does somebody have experimental analysis that indicates a numerical  
estimate of how many random transfers would produce how much  
fragmentation (extra entries) in the global route table for ARIN  
addresses?

John

On 2008Oct6, at 10:25 AM, Tom Vest wrote:

> ... To recap from last week, what makes routing keep
> working under the current paradigm?
>
> 1. CIDR  -- which provides the basic tools.
> 2. Top-level aggregation -- which the RIR community-system provided,
> and kept flexible over time as technology improved and RIR community
> practices changed.
> 3. Filtering -- which was/is only commercially feasible because of the
> top-level aggregation made possible by the RIR community-system.
> 4. Open entry for new routing service providers, which the arms-length
> RIR processes also enabled, and which effectively made aggregation and
> filtering "justifiable" and thus palatable to most direct
> stakeholders, as well as to the few indirect stakeholders/outside
> observer who knew that the system existed and understood the basics of
> how it worked.
>
> An uncoordinated market will eliminate (2), which will make (3,4)
> impossible, which will cause the current routing paradigm to fail in
> short order.
>
> Maybe a routing cartel will emerge in time to solve the problem,
> without creating new problems for aspiring new entrants and the ever
> widening audience of attentive external stakeholders. But that's a big
> leap of faith...
>
> TV

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20081006/7fa4feee/attachment.htm>