extensions: misbehaving cases

[vladopajic]

specification for gossipsub extensions, "test extension" and partial messages extension are not very well describing what should be considered as misbehavior (penalty).

for example:

• If a peer supports any extension, the Extensions control message MUST be included in the first message on the stream

  • if an extension control message is sent, but it is not the first message, is it misbehavior?

• Extensions control message MUST NOT be sent more than once

  • go-libp2p identifies this case as misbehavior

• A node MUST omit the partialMessage field of the PartialMessagesExtension message when sending a message to this peer. In other words, it MUST NOT send this peer encoded partialMessage data since it did not request it.

  • not clear if this should be misbehavior

in general, it would be better if specs had a better way of specifying what cases should be considered as misbehavior. for example, all documents (in gossipsub) could have an additional section where all misbehavior cases are listed.

---

edit: there are some cases when MUST is not use in spec. for example partialMessage rpc has groupID and topic fields which are mandatory for extensions to be set, but since they are optional (on protobuf) extensions should do senity checks on these fields and when those are not set, peer that sent message should be considered as misbehaving?

[MarcoPolo]

If a peer does not fulfill a MUST, it is not implementing the protocol.

Disconnecting from the peer and penalizing their local score is the correct approach. I'm happy to accept a patch that makes this clearer.

[Abidoyesimze]

Hi I'd like to work on this issue

Can you assign this to me @MarcoPolo