NFD

What's the use case?

faces/create command only supports UDP unicast and TCP unicast.
There is exactly one channel for each of udp4, udp6, tcp4, tcp6.

The use case we have in mind is creating multiple UDP faces towards the same remote endpoint, each of them using a separate local network interface. This requires allocating one channel per IP address, or at least one channel per interface, in the local NFD.

The use case described in note-2 has additional dependencies. Can Davide join an NFD call to discuss?

Indeed, several things are needed to obtain the goal described in note-2, this is one of them. If needed, we can open separate tickets for the others, as they affect different components of NFD.

Andrea and I can definitely join a call to discuss our use case and rough solution, it's entirely possible that we're attacking the problem from the wrong angle. When? Next Tuesday?

Also, let me clarify one thing here first. I realized that we don't really need/want to specify the local URI (e.g. in terms of local IP address), what we want is to specify which system network interface will be used by the created face, and then the face will keep using that interface even if its IP address changes. In a way, this would be similar to the -I option of ping(8).

Davide, this would have multiple complication, as it doesn't just depend on nfdc, but also on management protocol, face manager, protocol factory, and may be something else.

Do you really need to create multiple UDP faces to the same destination from the same NFD? Is it for some kind of debugging purposes?

Alex Afanasyev wrote:

Davide, this would have multiple complication, as it doesn't just depend on nfdc, but also on management protocol, face manager, protocol factory, and may be something else.

Yeah I'm aware of that. Andrea already has a (minimal and unpolished) implementation of almost all of the above for the UDP case. We're going to start using it in our V2I experiments asap.

Do you really need to create multiple UDP faces to the same destination from the same NFD?

Yes, we do. And not for debugging.

Davide Pesavento wrote:

Andrea and I can definitely join a call to discuss our use case and rough solution, it's entirely possible that we're attacking the problem from the wrong angle. When? Next Tuesday?

Ping?