[arin-ppml] Addressing for other planets

[Tony Li]

Hi John,


> If agencies are currently picking random IPv4 prefixes from their normal assignments, then that is indeed unfortunate – particularly since optimization of routing traffic will inevitably involve some carefully consideration of aggregation options and network architecture choices
> 
> (regardless of whether provider topological or celestial-body based.) 


Agreed, My understanding is that agencies operate much more like large enterprises than anything else.


>> We need this draft and corresponding RIR policies to take back to them and help them with aggregation.
> 
> It’s clear that some guidance is needed, but again, it appears that there are multiple approaches that such guidance could take; e.g.  establishment of a deep-space general purpose address range with provider-based allocations would also resolve the situation of agencies picking address space at random – and that is regardless of whether celestial-body topological considerations were incorporated into the accompanying allocation policy.


Agreed.  


>> That is correct.  Aggregation is sensitive to topology.  If the topology is provider-exclusive, then you would want to aggregate that way.  However, if the topology is provider-shared, then you would want to aggregate geographically.
> 
> Very challenging to perform any real optimization in advance absent some firmer planning assumptions, as either approach could easily end up being remarkably  sub-optimal.  


Well, the very good news is that even given optimal addressing, if our assumptions are incorrect, we lose nothing.  Better addressing enables better aggregation, but even if agencies are not willing to perform higher levels of aggregation, we will still have provider based aggregates.


> Are you suggesting that the community provide authoritative network architecture direction to the deep space agency networks?   (Yes, mandating a specific addressing plan which could result in highly suboptimal routing absent corresponding architectural decisions – i.e. celestial-body interconnections and transit – does indeed equate to providing binding architectural direction...)


Yes, we need to say that agencies should request allocations for deep-space purposes from the organization that is in charge of space allocations and that it should follow the policies of that organization.  Right now, there’s nothing in place to say that.

I’m not seeing that anything can end up in highly suboptimal routing if we at least get deep-space specific addressing in play.


>> Provider-based allocation allows for natural aggregation if there is a provider-based topology.  Provider-shared topologies break that natural aggregation.
> 
> 100% agreed - provider-shared topologies do have very distinct architectural and mission implications – but the mandate of the Internet Numbers Community to establish binding requirements of this nature (particularly on others not directly involved in the deliberations) certainly warrants some careful consideration by this community. 


Thus, this discussion…

Cheers,
Tony