NFD

we should be considerate to other implementations when doing such changes. In this case, the encoding should be added to the end (= the first thing to be prepended). This allows other implementation to successfully decode even if it is not yet supporting flags.

I disagree.
The previous and current version of FaceStatus both specifies a complete struct without allowing arbitrary TLVs at either the middle or the end.
A correct decoding procedure for the previous version of FaceStatus should have checked there isn't any junk at the end.
Thus, any management struct change would be a breaking change no matter what.

I disagree with your disagreement and will refer to the Postel's robustness principle.

Alex Afanasyev wrote:

I disagree with your disagreement and will refer to the Postel's robustness principle.

+1

While looking into the NFD portion of this task, I believe I found an issue with the code that was just merged. Flags should be added to FaceTraits, instead of FaceStatus, so that FaceManager::collectFaceProperties continues to work with both FaceStatus and FaceEventNotification (by accepting FaceTraits).

Eric, can you submit the fix asap, so it can be put in the soon to be released version of ndn-cxx?

Alex Afanasyev wrote:

Eric, can you submit the fix asap, so it can be put in the soon to be released version of ndn-cxx?

I'll work on it tonight. When is the release?

I've added Flags to the FaceEventNotification spec on the wiki.