0conf LN channels and v3 anchors

[MattCorallo]

@t-bast

I’m not sure that’s sufficient, because you usually want to allow having a chain of unconfirmed splices.

Mmm, that’s a good point, though that is still a relatively simpler chain that we could still analyze assuming no additional unconfirmed inputs, no?

@t-bast

The topologies that v3 would then need to support would probably be too complex to analyze before cluster mempool…

All of the v3 discussion assumes cluster mempool

@t-bast

My plan was to have a change output on the funding/splice transaction on B’s side to ensure that it could be CPFP-ed.

I think we should strive, if at all possible, to have splice bumps be RBF, not CPFP. Seems nuts to prefer CPFP, no?

@t-bast

This is indeed far from ideal, but that’s something we already need today if we want to protect against pinning, so v3 doesn’t make it worse?

Sure, but if we’re rethinking how we do mempool policy, we should probably think through these things so that we try to get as close as possible and don’t end up with a v4