[deepspace] Re: Understanding the difference in aggregation between a single and multiple RIR model

[Jen Linkova <[e-mail]>]

Hi Tony,

On Fri, Aug 28, 2026 at 10:31 AM Tony Li <[e-mail]> wrote:
> Ok, let’s try an example.  Let’s suppose that there is a future colony on Mars.  It is a joint project between NASA, JAXA, and ESA.  The colony is physically and topologically co-located, so for routing purposes, we would like it to be one prefix.
>
> In your scheme, you have three agencies going to three RIRs, each of whom makes an allocation.  I can see one of two things happening.
>
> In one case, the RIRs do not cooperate.  While they allocate from the block for Mars, they each allocate from their own private pools.  We effectively lose any aggegation for the colony.
>
> In the other case, the RIRs must closely cooperate.  They need a coordinated plan whereby they all collaboratively sit down, extract a prefix for the colony from the Mars block, and then subdivide that prefix to the three agencies.  Even the non-participating RIRs must be part of this coordination because the Mars block is a shared resource.  This seems like it has multiplied the administrative costs by a factor of five, for no benefit.
>
> Of these two schemes, I suspect that the RIRs are not going to be willing to absorb this amount of overhead, so they will end up choosing the first scheme.  Routing will suffer.
>
> I will also point out that your scheme requires that you approach all five RIRs and get them on-board with this policy.  This quintuples your work.
>
> In my scheme, one RIR is responsible for doing this allocation.  Done and dusted, very straightforward.

I'm sorry, I've not gotten my coffee yet, I need more help to
understand how things work in your example.
So 3 companies are building a joint colony on Mars. Are all 3 of them
going to the RIR, each asking for a netblock, and the RIR would
allocate 3 netblocks from a larger supernet? And all 3 netblocks will
be used for the colony network? What would happen if one company
decides to leave or 2 other companies join? More netblocks (not
necessarily aggregatable with the original one) would need to be
allocated, right? What happens if the new company, which has joined
the party, wants to use their addresses for their network? Is the
'BYOIP' model explicitly prohibited by the architecture?

Or do those 3 companies form a new legal entity which would become an
LIR and would ask for one netblock? In that case - in both your and
our proposal - it will be one LIR going to one RIR asking for a block.
That LIR would act as an ISP for whoever uses the colony network.

I'd really appreciate it if you could elaborate a bit. I feel like we
might not be on the same page regarding some implicit things..

> Repeat this scenario and you end up with no aggregation below the level of a celestial body.  IMHO, this is a nightmare.

There will be some aggregation, and, again, I'm not sure your example
leads to a fundamentally different.

> I remind you of Rekhter’s Law:  Addressing can follow topology, or topology can follow addressing.

Yes, let's choose ;)

-- 
Cheers, Jen Linkova