[ppml] ripe-55/presentations/bush-ipv6-transition.pdf

[Brian Dickson]

John Curran wrote:
>>> Randy wrote:
>>> and, given a choice of
>>>   o un/itu/... regulation
>>>   o national government regulation
>>>   o regulation by the rirs
>>>   o an open market
>>> i strongly suspect that the latter will produce as 'fair' a
>>> distribution of resources as any of the former.[...]
>>>       
> I believe that Randy's point is that an open market will likely
> result in availability of address blocks to all ISP's, both large
> and small, and this might not happen in some of the possible
> choices before us.
>
>
> /John
>
> p.s.  Of course, depending on the particular rules of the "open"
> market, there can be a higher probability of block fragmentation
> since there's more value in providing (n) sites with unique blocks
> than one site with a block (n) times larger.
Everything depends on the meaning and contect of "open".

Per Ben Edelman (who has been engaged as a consultant by ARIN and who 
spoke on IP markets),
getting into the details of how a market might operate, before deciding 
that a white market should
exist and/or be operated/facilitated by {ARIN|RIRs|IANA|whoever}, is 
putting the cart before the horse
(to paraphrase).

My $0.02 is:
I think the "open" in open market should be primarily limited to 
pricing. Anyone with address space should
be able to ask for whatever $$$ they think they can get.

However, I think that, since there is more than a mere commodity aspect 
to IP address blocks, that some
additional regulation would be needed to prevent disasterous outcomes.

Things to avoid:
- market manipulation by collusion, cartels, or any other kind of 
deliberate action
- direct sales between parties (this is something better characterized 
as "grey" or "black" market)
- speculation, cornering of the market, or other blanket market abuse by 
those with deep pockets
- purchases directly by recipients, without approval or intermediation 
by RIRs or similar agents (to prevent inappropriate allocation of resources)
- deaggregation

Things to encourage:
- transparency of operation of the market
- blind, double-blind, or N-tuple blind market
- separation of the two sides of the market, i.e. those selling blocks 
and those buying, by RIR-type entity (who is also "blind")
- aggregation between return of blocks to RIR-type entity, and 
allocation/sale to RIR customers
- decoupling buying/selling aspects by way of combined purchasing pool
- decoupling funding of purchases from allocation (effectively turning 
"purchasing" into "donating to the fund for getting space back into the 
RIR pool")
- creating contingency-blocking "trees" similar to real-estate 
"chains"-of-conditional-purchases, e.g. in real-estate, there can be 
multiple transactions which
  are effectively blocked, until a transaction which has no conditions 
on it, is able to trigger the completion of all of the transactions in 
the chain. This can be
  extended with a non-trivial dependency topology, where N:1 and 1:N 
branching/merging of transaction sets, trigger "go/no-go" events on the 
whole tree.
- coupling contingency-based bidding/donating with purchase&allocate 
events - someone's donation gets accepted iff their allocation request 
is satisfied

It's a highly non-trivial kind of thing to delve into, but IMHO 
something which can work very well, *and* actually has some chance to 
improve the aggregation
situation in IPv4.

Imagine getting some return on the effort involved in renumbering, when 
previously the only benefit was knowing you were doing "the right thing".

Brian