[address-policy-wg] ALLOCATED PI / UNSPECIFIED next action

Hi,

On Thu, 20 Oct 2016, Ciprian Nica wrote:

> I agree with Daniel. A well defined problem is half of the solution.

+1.

> In this particular case the problem arises because the main question is who makes the money, the LIR or the end user.

And when noone is really "making money", but is just using the number 
resources as they were originally distributed...???


> In the past there have been PAs used as PIs

Yes. But that doesn't qualify as "a mess" or "swamp"?


> so technically I think the "allocated" part should be the one that's 
> more important, therefore I would support the replacement of allocated 
> pi & unspecified to allocated pa.

Why would you want to change status of something that belongs/is in use by 
other organizations?
With or without agreement from the end user (and LIR)?


> If you ignore the greed then changing the status would not make any 
> difference. 

Maybe we need to revisit how this "issue" was created in the first 
place... (and when).


> Ripe has a relation with the LIR and the LIR with the customer.

1st part is true. 2nd part i really wish it was true... this (ALLOCATED 
PI/UNSPECIFIED blocks) is not LEGACY space, but there are some tricky 
details too... :-(


> Changing PI to PA will not affect the workability of the IPs nor the 
> relations that are already in place.

Yes they can be..... and this can probably be observed from the routing 
info.

Example: LIR manages /16, everything is PI, so every /24 is 
already "globally routable". Once that /16 automagically becomes PA, 
anyone in the world can happilly reject any /24 from the /16. This is 
even harder if end users are not using the same upstream and/or don't 
even have any relationship with the LIR -- can happen, does happen because 
this goes wayyyyyy back.

Please keep in mind this still comes from the 20th century... ;-)


> Changing them to regular pi assignments would break the link between 
> lir and customer and give the enduser a possibility to make money, 
> nothing more.

???
Afaik, they are already PI. The link is (probably) mostly broken, and, 
imho that's the core of the "monster" we have to deal with ;-)


Cheers,
Carlos


> Ciprian
> 
> On Thursday, October 20, 2016, Daniel Stolpe <stolpe at resilans.se> wrote:
>
>       Thanks for the update and summary Sander!
>
>       I have been thinking a bit both during this particular case and in general about something from another working group. Job introduced the concept of
>       "numbered work items" with several phases and where the first phase reads like (quote):
>
>       phase 1: problem definition
>           In this phase as group we'll work on formulating an exact problem
>           definition: text goes back and forth in the working group, example
>           cases of the problem are provided. In a 2 or 3 weeks timeframe the
>           chairs declare consensus on the problem statement of NWI.
>
>           phase 1 output: clearly defined problem statement, or a conclusion
>           we cannot agree upon a problem statement definition. If the latter
>           is true, the NWI cannot proceed to phase 2.
>
>       (end quote).
>
>       Maybe it is only me but I have had the feeling sometimes that we are not completely sure what problem we are trying to solve, and that we can sometimes
>       start with proposing a solution before the problem is well defined or agreed on.
>
>       I think I am correct that the problem in the particualar case is unclear (unspecified or pi that is maybe not really pi) and/or incorrect (pi that is
>       really pa) data and according to that I agree with Sanders summary below.
>
>       Cheers,
>       Daniel
>
>       On Thu, 20 Oct 2016, Sander Steffann wrote:
>
>             Hello working group,
>
>             The discussion on how the RIPE NCC should deal with ALLOCATED PI / ALLOCATED UNSPECIFIED has died down a couple of weeks ago. We therefore
>             think that it is time to draw conclusions.
>
>             A total of 16 people and the working group chairs participated in the discussion following Ingrid?s proposal on how to handle the situation
>             of PI assignments within ALLOCATED PI / UNSPECIFIED. Five people (Sergey Myasoedov, Radu-Adrian FEURDEAN, Nick Hilliard, Leo Vegoda and Hank
>             Nussbacher) created side-threads without expressing an explicit opinion on the proposal.
>
>             The remaining 11 people were:
>             ?    Patrick Velder
>             ?    Larisa Yurkina
>             ?    Randy Bush
>             ?    Enno Rey
>             ?    Herve Clement
>             ?    Stefan Schiele
>             ?    Markus Weber
>             ?    Carlos Friacas
>             ?    Leo Vegoda
>             ?    Andre Chapuis
>             ?    Daniel Stolpe
>             ?    Oliver Bartels
>
>             Four participants stated that they represent organisations holding such allocations (Larisa, Markus, Andre and Daniel).
>             Three people indicate that they are related to PI assignments within such allocations (Enno, Stefan and Oliver).
>
>             Five people stated their clear support for the proposal (Enno, Stefan, Oliver, Patrick and Herve), mainly to increase clarity for PI
>             assignment user and to support correct registration.
>
>             While there was no explicit opposition, Larisa and Andre stated that it would create extra workload for their organisations while they don?t
>             really see the gains of such change. Larisa suggested to introduce alternative RIPE database statuses instead.
>
>             The other participants had mixed opinions:
>             Markus understands the advantages for PI assignment users, but was concerned about the extra workload for his organisation. He suggested to
>             somehow lock PI?s within the allocations and force the PI holders to sign contracts, but recognized that this idea might be not practicable.
>             Daniel could life with the change from ASSIGNED PI to ASSIGNED PA but agreed with Larisa and Andre that there is actually no issue to be
>             fixed.
>             Randy supported the aim of correct registration but also stated his concerns about the routing table and that some PI holders might not be
>             happy to pay a fee for the sponsoring LIR.
>             Carlos also stated his concerns for the routing table.
>
>             Conclusion:
>             Five people supported the proposed approach, four people saw some advantages but also were concerned about side effects, while two people
>             didn?t see the need to take action.
>
>             There were three opposing arguments:
>             - big workload compared to the gain
>             - increase of the routing table
>             - PI holders might not like to pay a fee for the sponsorship
>
>             The first opposing argument can be considered as addressed as three PI users confirmed that a clarification of that issue would be very
>             important to them. And the RIPE NCC can support the LIRs, for example making bulk updates on route and domain objects.
>
>             The second opposing argument, could be considered that this is not directly related to the fixing of the registration. Already now all but
>             one of the allocations in question contain more specific route advertisements. Also in the extem case that all ASSIGNED PI within the
>             allocations would be carved out, we would talk few thousand new entries in regards to 628K total routing entries (normal growth of the
>             routing table is around 2K per week).
>
>             The third opposing argument was addressed by Gert, stating that PI holders appreciate to pay a small fee to be sure that their resources are
>             correctly registered.
>
>             Based on all of this I feel we have a strong enough mandate for the RIPE NCC to move forward, but some concerns about the amount of work
>             involved. I therefore would like to ask the RIPE NCC on behalf of this working group to move forward with their plan, but to extend the
>             proposed deadline of the end of January 2017 by a few months (the end of Q1 2017) to give LIRs a little bit more time if needed.
>
>             Cheers,
>             Sander
>             APWG co-chair
> 
>
>       _________________________________________________________________________________
>       Daniel Stolpe           Tel:  08 - 688 11 81                   stolpe at resilans.se
>       Resilans AB             Fax:  08 - 55 00 21 63            http://www.resilans.se/
>       Box 45 094                                                            556741-1193
>       104 30 Stockholm
> 
> 
> 
>