Re: [sig-policy] Prop 050(072) comments
Terence Zhang Yinghao said the following on 20/3/09 11:26:
>
> My concern about prop-050 is mainly on the process and the time of implement,
> I feel now that we still have free pool to allocate, better not to implement
> a policy which may encourage the address market;
But there is a difference between preparing a policy for implementation,
and actually implementing it. That's why I was asking how long you think
we need to discuss the policy proposal so that it would be ready to be
implemented. Sam has just replied saying that implementation might need
6 months after EC endorsement. So that's a minimum of 6 months after
reaching consensus, which means consensus has to be achieved at least 6
months before we need to implement the policy.
What I thought we were discussing right now is a policy that we can
implement when it is required.
Are you suggesting that we don't even discuss what we should implement
until we no longer have a free pool to allocate from?
> To answer some of your questions regarding CNNIC, our members do have some
> discussion on the transfer proposal, and their views are summurized as following:
Are these recent views, following the APNIC meeting in Manila? If they
were accumulated over the last two years, it's a great pity they
couldn't have been shared with the entire community from when Geoff
first brought the transfer issue to our attention.
> Arguments favor the transfer:
>
> --Once the fresh IPv4 allocations pool run out, and IPv6 deployment is not ready,
> a mechanism is needed to re-use of the allocated but unused IPv4 resources.
Did they have suggestions as to what this mechanism might be? Mechanisms
for reclaiming unused but allocated or assigned address space have been
tested ever since the Internet moved from classful to classless routing,
and I wouldn't say they were that successful. Saying a mechanism is
needed is fine, but what is it?
What we do know, from the real world out there, is that when there is a
shortage of a resource, people will do anything to obtain some more. It
might not have been the case in China, but when the price of oil was
around US$150 per barrel, the price of petrol and diesel reflected that
extreme price. There were shortages at fuel filling stations, large
queues, and very high prices in many countries around the world.
> Arguments against the transfer:
>
> --Address transfer contradict to the current need based allocation policy
I think we are all in agreement here, and have been probably for the
last 2 years this has been discussed. But it doesn't help us with
dealing with the issue that people out there are actually transferring
addresses already...
> --It attaches a potential 'value' to IP addresses, and may attract some businesses to
> apply for more IP addresses than their actual need, there for speed up the
> IPv4 addresses consumption.
IPv4 addresses have had a "value" for a long time. Companies that
collapsed in the dotcom bust in 2000/1 had their assets taken over by
other companies. These assets included the IPv4 address allocations used
for existing networks which were taken over by the purchaser.
Is there a problem with a speed up of IPv4 address consumption? We only
need to shout "petrol shortage" on the radio or television and people
will queue up at filling stations to fill up - even though they probably
don't need to and there is no real shortage. We more than likely will
see a speed up of IPv4 address consumption anyway, whether there is a
transfer policy or not.
> --It is discriminatory to LIRs in developing countries who have fewer IPv4 resources
> than other countries, as they have to pay more to get the addresses they need.
As long as there is a free pool of IPv4 addresses, everyone can go to
their RIR and get address space. When there is no longer any free IPv4
addresses around, people will be swapping addresses for money, whether
APNIC has a transfer policy or not. And when they pay for address space,
it discriminates against those who cannot afford the huge rates that
will be charged. And that won't be just developing countries either! ;-)
> --It implicitly recognize the market of transfer and may encourage the
> tranfer market and may change the way IPv4 addresses currently managed.
That's very possible, I agree. What would the CNNIC members suggest
would be a better alternative? Transfers are happening right now, like
it or not, so we need to find a responsible way of handling the
situation. Hence the discussion that's been ongoing for the last 2 years
or so. :-)
> --It may deaggregate the address block and lead to rapid growth of the
> routing table.
Of course it will. But given that many ISPs are wilfully deaggregating
anyway, with impunity, this is just more noise in an already noisy
routing system. Take a look at www.cidr-report.org for some stunning
examples of existing badness.
> Generally speaking, we all agree that keep a accurate record is important, but
> most of our members don't insist on we must have a transfer policy or we should not
> have a transfer policy. They pay more attention to if there are negative impacts and
> if those impacts can be mitigated to an acceptable level.
Any suggestions on how to keep an accurate record while dealing with the
situation where some organisations are transferring address space to
each other? And how to avoid abusing the system?
Because that's what's been keeping us quite busy for a long time now.
It's fine to propose conditions, but they do need to be backed up with
what the details might be, so I would definitely welcome suggestions.
> We have very few case of IPv4 address transfer in China though, till end of
> 2008, there are about 298 million internet user in China main land, annual
> growth rate is about 41%, we only have about 181 million IPv4 addresses, annual
> growth rate is about 34%, so I think there are very few unused address blocks for transfer.
Very few cases is not the same as zero cases. Given that transfers
currently are not allowed under APNIC policy, I assume that CNNIC went
to these organisations and successfully retrieved the transferred
addresses for reuse? If CNNIC didn't do this, what did you do, and how
can you justify that action when there is no policy to cover it?
> Serveral of the tier-1 ISP in China own large portion of the
> IPv4 allocation, when customer no longer use their services, they just get the
> addresses back and assign to new custmer when needed, if we can call this
> 'transfer',that's the way it happen.
No, that's not a transfer. That's the proper operation of PA space.
If one of the CNNIC members received address space from you, decided
they didn't need it and "gave" it to someone else to use rather than
give it back to you, that's the type of transfer we are talking about here.
philip
--