[address-policy-wg] 2012-05 New Draft Document and Impact Analysis Published (Transparency in Address Block Transfers)

* Milton L Mueller

>> I object to publishing information of rejected transfers (and, by 
>> extension, rejected pre-approvals). The NCC does not publish any 
>> information about rejected PA allocation requests either, and I
>> don't see why transfers should be any different.
> 
> [Milton L Mueller] You have not given us any reason not to provide
> this information, other than "we haven't done it before" - which I do
> not accept as a good reason.

I, like the NCC, consider this data to be confidential.

Let's say I want to transfer an allocation to another LIR, and want the
fact that I've been dealing with said LIR to remain a secret until the
deal is done. If the deal falls through due to the failure of the
recipient LIR to justify their need for the transferred resource, I
don't want the fact I was in negotiations to transfer away 192.0.2.0/24
to become public knowledge.

> I will tell you why we need to do this. There are serious concerns in
> the existing transfer market about potential discrimination in the
> application of needs assessments, at least in the ARIN region. Needs
> assessment is not an entirely scientific or objective exercise.
> Providing basic statistical information about how many applications
> are rejected avoids undermining confidentiality but also provides
> some knowledge as to how many attempted transfers are coming in and
> how many are rejected. If there are complaints about the application
> of needs assessment criteria - and there already are in other regions
> - at least the community has some information about

This is a red herring. The proposal does not call for «basic statistical
information about how many applications are rejected», it calls for the
explicit identification of origin LIR and the address block it intended
to transfer away.

Also, if the goal of the identification is to combat discrimination in
need assessment, isn't it the *receiving* LIR that should be identified?
How is the identity of the LIR giving up its allocation, and the
identity of the allocation itself, relevant to need assessment?

I have no objection to having the NCC publish aggregate information
about how many transfer tickets they've handled and how many of them was
approved. That said, I'm not so sure we would need to have a policy for
it, it might be simpler to just ask them to publish the information in
aggregate form. For example, they do something along those lines at
http://www.ripe.net/lir-services/resource-management/tools-for-lirs/reponse-time-ipv4
without there being any policy compelling them to.

>> That said, I am not quite sure it is really necessary, as all the
>> requested information (except rejections) is already available:
>> All allocations are listed in alloclist.txt along with their date
>> and holding LIR (reg-id and name). It will be trivial to check for
>> transfers - allocations made after the last /8 policy comes into
>> effect outside of
> 
> [Milton L Mueller] Trivial to whom? To someone who has a system set
> up to constantly track this or who has bulk access to the entire
> database and can write scripts to check for it on a regular basis?

Yes. It's not difficult. You basically download two files and see what
has changed.

> This is not reasonable. RIRs need to recognize the significance of
> the emergence of a transfer market. The attitudes here in RIPE that I
> detect are incredibly complacent. A market for addresses will
> profoundly transform the nature of IP number allocation. All RIRs are
> obligated to do everything they can to make this market easily
> understood, transparent and efficient. Why shouldn't they? Can you
> provide a positive reason why not?

I think both the RIPE community and the NCC fully realise that transfers
will play a significant role in the near future.

> [Milton L Mueller] I think the description you provide of what would
> be necessary to track this information without the policy proposed in
> 2012-5 self-evidently proves that this policy is needed. Your
> approach is complex, expensive and would limit this information to a
> few specialists who would keep the data proprietary in order to
> protect their investment in collecting it.

I think you are greatly over-estimating how difficult it is to parse and
extract information from the stats files available on the FTP. I think
anyone with basic programming or scripting skills would have no
problems. I'll be happy to provide you or anyone else with a script that
generates the report you want based off the stats files, if the NCC
makes all the necessary information available there.

> My proposal would make it
> accessible to anyone. You have not provided any good reason to limit
> it in that way. You have not provided any costs or harms that would
> occur if the information is made accessible. Perhaps you can do so?

Like I said, I have no objection to the proposal other than the part
regarding publishing identifying details about rejected transfers.

That said, I think there is likely much faster ways to accomplish the
desired transparency than going through the PDP. I think going through
the PDP is over-engineering. Don't mistake that for opposition to the
proposal though.

>> I think that a simpler and probably much faster way (no PDP!) to 
>> accomplish the desired <transparency in address block transfers>
>> would be to simply ask the NCC to publish historic versions of
>> alloclist.txt, and/or to include a regid/LIR column for resources
>> in delegated-ripencc- extended-latest (for which historic archives
>> are available). This has been recently suggested in the services
>> wg, but it was objected to, because of privacy issues for
>> assignment holders (so irrelevant relevant to this proposal). I
>> think I'll go and revive that thread now...
> 
> [Milton L Mueller] No, that is not "simpler". Not by any reasonable
> definition of "simple." RIPE could easily make this accessible to all
> with a few keystrokes and procedural change at the source. Can you
> explain why it should not?

>From the impact analysis: «It is worth mentioning that the RIPE NCC is
willing to publish details on resource transfers and a report has not
been requested so far».

So yes indeed, «RIPE[sic] could easily make this accessible to all with
a few keystrokes». They have stated a willingness to do so, too. So why
do we need to change policy, exactly? The PDP is a slow process. It
seems to me that it is faster to just ask the NCC to start publishing
the desired information. If they refuse to do so, then let's look into
compelling them through policy.

IMHO,
-- 
Tore Anderson
Redpill Linpro AS - http://www.redpill-linpro.com