Yes, spec said use channel_types if and only if both sides send them. proposed a feature bit as well, but that's strictly unnecessary: just to show up-front you support it. But it seems this caused confusion?
Generally in future we should add a feature bit IMHO.
Conversation
Still not sure why we need to bother? lnd having a bug implementing the spec isn’t a bug in the spec. I understand lnd has a bunch of infrastructure built up around feature bits for everything but making the rest of us do the same just for them seems like the wrong way?
2
5
Personally I’d prefer feature bits for ~nothing unless you need to find peers based on support (which is rare) or need to signal a single bit and it’s the most logical place. Adding a second bit (beyond presence of a TLV) to negotiate features is just complication for no reason?
2
2
That's how I view it as well. If we put feature bits for everything, it feels like we're not taking advantage of the "it's ok to be odd" rule
1
3
Sure, but if this is going to become *compulsory* the feature story is a bit neater: eventually, everyone supports it and if you don't you get an immediate error from your node when connecting, rather than some remote error msg when you open.
2
3
+1, I think the odd rule and feature bits in most places are compatible, feature bits will let us eventually *only* use explicit funding which is superior from an API perspective imo: the user *knows* which channel they'll get, no surprises
1
the other thing is that in a sense we were all *somewhat* correct, eclair+CL deployed before that latest PR (adding the feature bit, which was one source of divergence), while lnd deployed afterwards, and now the "most recent compliance" is that commit effectively
1
at least with feature bits I can check a nodes advertised set and see "how recent" they are, vs some interactive dance to see if they set certain TLV bits, etc
not saying there's some simple blanket answer here though, but def more stuff for us to dig into/clarify
1
> feature bits will let us eventually *only* use explicit funding
how does the feature bit change your ability to do this. I'm still just really unclear on what it changes from a UX PoV, and arguably just adds more code to get to the same place a second earlier.
1
> lnd deployed afterwards, and now the "most recent compliance" is that commit effectively
which PR are you referring to? If you mean github.com/lightning/bolt that isn't merged yet - and it seems you're not compatible with the latest version of that either?
But, yes, twitter isn't really a good venue for this - its a deeply philosophical debate. That said I'm not sure we're gonna agree on all the details in any case - to some extent these things are very case-specific, with different reasoning per feature.
2