> An interesting point. There's a whole body of work in "Disruption
> Tolerant Networking" to send packets (which they call bundles) from a
> node to another node, without assuming e2e connectivity. But, I am not
> aware of routing that really works in the general case. There is
> definitely a tradeoff of reliability vs how much you replicate.
>
> https://tools.ietf.org/html/rfc4838
> https://tools.ietf.org/html/rfc5050
>
> But, I think what you are suggesitng is simpler, because it's about
> replicas in the connected subset, not the full disconnected case.
Right, I wasn't thinking of dealing with partitioning. Not sure if
partitioning is important to handle, nor even if it makes things
really harder.
Anyway, I have a feeling we're fairly offtopic by now.
Stefan
