[address-policy-wg] 2013-03 New Draft Document and Impact Analysis Published (No Need - Post-Depletion Reality Adjustment and Clean up)

* Filiz Yilmaz
> My observation is that the new document proposed is not exactly a
> "registration" policy document.
> It more looked to me like a description of how address space management
> is done within RIPE NCC region "by the RIPE NCC". 
> 
> If I am not missing anything crucial, the main points described are:
>  
> - The language is English, 
> - How big an allocation can be, 
> - That there is no PI anymore to be assigned directly from the NCC pools
> to End users (except for the IXPs) so all resources will only goto LIRs 
> - And these blocks can be transferred between LIRs.
> 
> The only bit about registration I see in the new text is section 4.0
> Registration Requirements and it does not go more than saying details
> should be recorded in the database. 
> 
> So it does not contain any substantial information for registration or
> address management on the LIR's side.

Hello Filiz,

The proposal is a modification of the existing policy document
(ripe-592), it is not a brand new document written from scratch. The
language regarding registration requirements is not changed, so the
proposal does not change anything - it merely upholds the status quo.

We can certainly discuss amending the current registration requirements,
but to me this topic does not seem germane to the discussion of 2013-03.

> This is interesting as now with this proposed policy any End user's
> chance to get any IPv4 address space will be through an LIR and hope
> that these LIRs are responsible and know what they are doing. I would
> like to see some guidelines or at least principles mentioned in this
> document so the LIRs know their responsibility in terms of fair address
> management as well as the End Users so they know what to expect from
> these LIRs. This is what I would be expecting from a transparent
> documentation of a set of policies and principles that are still in place.
> 
> We may not have too many specific policies to set for the few left-over
> resources but I would like to believe we still have "principles" towards
> the responsible management of these resources. 
> 
> In that sense Michele has a point and I argue that LIRs need to be
> guided for "good address management" even without the "conservation"
> principle as the top priority in the new IP world. This is missing in
> the proposed policy text for it to be considered as a helpful
> "registration" policy in my opinion.

I have no problems being sympathetic towards a "good address management"
principle, but the simple fact is that this is not written down in our
current policy. What constitutes "good address management" is something
left to the LIRs to decide for themselves. Example: As a LIR, you are
today free to assign your last free IPv4 block to a group of spammers at
the expense of having to turn down the Red Cross, perhaps because the
spammers were willing to pay you more. Is that "good"? I'd say quite the
opposite, but yet it is perfectly allowed by today's policy.

So while I'm all for discussing the codification of a set of principles
in our policies telling our community to be "good", I'd rather not weigh
down this proposal with the addition of a brand-new "principles"
section, which I suspect would require a lot of back-and-forth
word-smithing to make everyone happy. So if we do want such a section,
let's add it in a separate proposal.

> In practice, I can set up a new LIR now and ask for a new allocation and
> I may be someone who does not have any previous RIPE or RIPE NCC
> experience. 
> If all I have is this document, I am not sure if it tells me enough
> about my responsibilities, while I will be a critical token in the EU
> address management and registration system by just becoming an LIR.

As a new LIR, all the IPv4 address space you are going to get from the
RIPE NCC is a /22, or 1024 addresses - hardly a «critical token in the
EU address management and registration system» by any stretch of the
imagination. This is the status quo today, and this proposal merely
upholds it.

I also find it rather hard to believe that an organisation willingly
joins the NCC to become a new LIR, acquires an IPv6 allocation, and
finally requests its first and only IPv4 /22 - only to find itself
confused about how to go about using it in a sensible manner and end up
assigning it away to an end user without any actual operational need.

>   * As mentioned in previous sections, the policy proposal would
>     negatively affect the ability of LIRs to engage in inter-RIR
>     transfers, as the RIPE NCC’s service region would be the only one
>     without a needs-based requirement for transfers.

I feel this statement is somewhat disingenuous. The "odd RIR out" here
is really ARIN, who is the only one to have a inter-RIR transfer policy
that has a needs-based requirement that is also applied to the other
region involved in the transfer.

The RIPE region does not have a inter-RIR transfer policy *at all*. The
same goes for LACNIC and AfriNIC, AFAIK. So as things stand today,
2013-03 cannot possibly "negatively affect the ability of LIRs to engage
in inter-RIR transfers" because such transfers are completely verboten
in the first place.

That said, there has been an inter-RIR proposal (2012-02) on the table
for a while now, but the community does not seem to want or care much
about it. I think that this is important to keep that in mind when
deciding on how much weight to give this argument against 2013-03.

APNIC does have a inter-RIR transfer policy, but it does not require the
other region to have a needs-based requirement. It explicitly states in
section 4.3 that for transfers where the recipient is in another region,
any conditions on the recipient is up to the recipient's region to
define. So (assuming for a second that 2012-02 has passed) 2013-03 will
not have any negative impact on transfers between the APNIC and RIPE
regions.

Another quite interesting thing to consider here is that APNIC initially
had a transfer policy that did not require any need evaluation. It was
only after the ARIN community changed their inter-RIR policy to
explicitly require the other region to have "needs-based" policies APNIC
added the need evaluation requirement to their general transfer policy.
I think it is interesting to consider if yielding to the ARIN demand
would actually be worth it, and APNIC actually provided some relevant
data in this regard - on May 15th 2013, one of their reps explained to
the APWG session at RIPE66 that there had been 11 inter-RIR transfers
between ARIN and APNIC. APNIC depleted their IPv4 pool on the 19th of
April 2011. Contrast this to the number of assignments that are being
made in our region: The NCC reported that between depletion on the 14th
of September 2012 and May 10th 2013, 251,254 assignments had been
registered in the RIPE database.

If we do linear extrapolation of the above to get "per year" numbers,
what this boils down to is that we ask the entire RIPE community to fill
out, validate, and archive 386,327 ripe-583 forms per year to allow 5
LIRs to engage in inter-RIR transfers with the ARIN region. You be the
judge if this passes your idea of a basic cost-benefit analysis. It
certainly doesn't pass mine.

I'm not at all principally opposed to inter-RIR transfers, so if the
author of 2012-02 would add a "need compatibility clause" to the
proposed inter-RIR policy, I would have nothing against that. In other
words, if it would be possible to make the need evaluation a voluntary
thing that you only had to do if transferring addresses from the ARIN
region - fine. Then those 5 LIRs, and those 5 only, could subject
themselves to whatever demands the ARIN community might have, while the
rest of us would be free of the IPv4 bureaucracy. Win-win.

>   * Implementation of the policy could expose LIRs to legal challenges
>     under EU competition law.

If an LIR is willing to engage in unlawful anti-competitive practises,
it will of course be exposed to legal challenges. This is certainly
nothing new brought on by 2013-03 - the law trumps *any* address policy
we can produce here, and that's how it's always been and always will be.

Furthermore, it makes absolutely no sense to me for the community to
demand that every single LIR and End User in the community uphold the
"need" bureaucracy in an attempt to somehow shield or indemnify "bad"
members of the community engaged in anti-competitive practises from
legal repercussions.

> I think singling out the RIPE NCC region in the world of transfers may
> not be the best idea at this stage.

This "world of transfers" has failed to materialise. The actual
statistics show that the second-hand IPv4 transfer market is at most
supplying 3-4% of last year's (pre-depletion) demand in our region. If
we had completely removed the entire transfer policy today, I think that
as far as the internet is concerned, tomorrow would have been business
as usual. It seems that for most folks dealing with IPv4 depletion, CGN
is the solution, not transfers.

I think Rob Blokzijl hit the nail straight on the head when he in the
RIPE66 APWG session warned against getting bogged down with discussing
the ins and outs of «little add-on policies like transfer». After all,
this proposal is a large and rough clean-up - it will be much easier to
fine-tune and polish the remaining policy afterwards (transfers, PI/PA
merging, etc.). Like I've said before, this proposal is essentially the
amputation of the policy limbs that died following IPv4 depletion, and
any following cosmetic surgery is best left for later.

Best regards,
Tore Anderson