Add option to allow even custom messages to be sent

[benthecarman]

Closes #4960

Adds an option --experimental-all-onion-messages that will behave like --experimental-onion-messages but also allow even numbered messages to be sent

[ZmnSCPxj]

Mostly OK, suggest expanding the err message though.

[benthecarman]

Renamed from --experimental-even-onion-messages to --experimental-all-onion-messages, I think this is more clear

[vincenzopalazzo]

LGTM from a code point of view, but I'm a little bit worried about the dropping connection in some cases.

How you can be sure that there is no problem with your communication?

BTW, I have no enough knowledge to judge this PR from a correctness point of view

[cdecker]

Can you write a test to verify that even messages are indeed forwarded to the plugin if enabled and drop the connection otherwise?

It'd also be helpful to get some rationale for the change. Are you working on a new sub-protocol or is this to build a plugin for compatibility with an existing sub-protocol?

[benthecarman]

Can you write a test to verify that even messages are indeed forwarded to the plugin if enabled and drop the connection otherwise

Yes

It'd also be helpful to get some rationale for the change. Are you working on a new sub-protocol or is this to build a plugin for compatibility with an existing sub-protocol?

Yeah, I outlined in #4960 that DLCs use even numbered types for messaging and am planning on creating a plugin to send and receive the DLC messages.

[cdecker]

Excellent, just saw the original issue now, that makes sense to me. Sounds like a good experimental change, with no promise that it'll graduate, but the experimental deployment will inform us about the safety.

[benthecarman]

Added a test but in places like here, the message is dropped. I am trying to use the experimental-accept-extra-tlv-types config, but don't know how to access it from channeld, openingd, closingd, etc, is there an easy way?

[rustyrussell]

Hmm, this will get much easier once my PRs drop which move this handling to connectd. That should be in the next week; can this wait?

[benthecarman]

Hmm, this will get much easier once my PRs drop which move this handling to connectd. That should be in the next week; can this wait?

Yes, can you link them so I can rebase on top?

[rustyrussell]

First draft at #4985...

[benthecarman]

Hmm, this will get much easier once my PRs drop which move this handling to connectd. That should be in the next week; can this wait?

Rebased on master, don't see how I have access to allow_even_custom_message inside of multiplex.c.

Is this something someone could help me with?

[rustyrussell]

Makes sense to allow everything through here, but be careful: we should change the semantics of custommsg hook to allow it to reply "result": "accept" (and terminate traversal of hooks if it does). If nobody does this for an even message, we should tell connectd to send an error and close.

[niftynei]

@benthecarman we're looking to get an RC out this week, any chance you'll get a rebase / changes in in the next few days? If not I'll mark it for the next release.

[benthecarman]

@benthecarman we're looking to get an RC out this week, any chance you'll get a rebase / changes in in the next few days? If not I'll mark it for the next release.

Hey sorry I won't have time in the next week or two. Last I touched this I wasn't really sure what to do next

[cdecker]

Removing milestone target for now, as this requires some work. Feel free to undraft as soon as you'd like me to re-review and assign to a milestone.