Recommend that PATH_ABANDON should be retransmitted on another path

[mirjak]

Maybe this is obvious but if a path_abandon frame gets lost, this will usually lead to a PTO. In this case the path is highly likely to be already broken and I probably should retransmit the path_abandon frame on another path to actually be able to close the path. We already say in the new "handling PTO" section that it makes sense to generally retransmit on another path. However, maybe we should actually say this using normative language (SHOULD or MUST?) in the path closure section in case of the path_abandon frame?

[huitema]

The feedback from Dublin was quite clear: a bunch of these strategies are still research issues. We should generally not try to solve them as part of the draft, because it makes the whole draft look like a research project. And we should certainly not make normative statements that implementation SHOULD or MUST implement a variant of an algorithm for which research is still ongoing.

[mirjak]

Actually we already have this sentence in section 3.3 (Path Closure):

Thus, even if connectivity on that path is already broken but there is still another usable path, it is RECOMMENDED to send the PATH_ABANDON frames on another path.

I would maybe just like to add one more sentence here to make the case above a bit more clear like this:

Thus, even if connectivity on that path is already broken but there is still another usable path, it is RECOMMENDED to send the PATH_ABANDON frames on another path. For example if the PATH_ABANDON frame itself got lost on the to be abandoned path, this might be a sign that the path is not usable anymore and retransmission on another path would address this concern.

[mirjak]

I created PR #473 to see the proposed change in context.