Rethink default dust limit

[TheBlueMatt]

It seems the current dust limit (5k * current feerate) is low enough that with a 1sat/vB feerate we may not allow any dust HTLCs at all. Someone should run the math more carefully and we should pick a new default.

[tnull]

Isn't this mainly an issue on Signet? Should we start introducing more per-network defaults/settings which would, e.g., also allow us to introduce some (~static) defaults for Testnet? This might improve the interop situation with LND and CLN, which (AFAIK both) set static fee rates on testnet that regularly conflict with our dynamic estimates.

[TheBlueMatt]

Hmm, I mean if feerates ever drop back down to 1 sat/vB on mainnet it would be an issue there as well, I assume. It may be that there's no reasonable value here that scales across feerates (and maybe we need a multiplier + const config, rather than a simple multiplier), but it would be good to run some analysis here that's more rigorous.

[TheBlueMatt]

As for static feerates on testnet, I think we can just do that downstream of LDK? Maybe document it in the estimator? I guess we could add wrappers in Channel when looking at feerates to check if the network is test and accept anything? That's a separate issue, though, IMO.

[tnull]

As for static feerates on testnet, I think we can just do that downstream of LDK? Maybe document it in the estimator? I guess we could add wrappers in Channel when looking at feerates to check if the network is test and accept anything? That's a separate issue, though, IMO.

Yeah, might be a separate issue. But, I think it would be good to either automatically set defaults compatible with what LND/CLN is doing or to at least provide very clear guidance in the docs somewhere, because users keep running into these issues on testnet.

[TheBlueMatt]

Can you open an issue to track it (maybe with some color on how often we see this and in exactly which ways)?

[tnull]

Can you open an issue to track it (maybe with some color on how often we see this and in exactly which ways)?

Done: #2925