Lightning Specification Meeting 2023/02/13

[t-bast]

Lightning Specification Meeting 2023/02/13

[t-bast]

Dual-funding cross-compatibility has been successfully tested between eclair and cln 🎉 🎉
A bit more details are available here: #851 (review)

[carlaKC]

Transcript: bitcointranscripts/bitcointranscripts#232

[t-bast]

I won't be able to attend this meeting, but I will report on cross-compatibility test results between cln and eclair for dual funding before the meeting starts.

[joostjager]

Highly Available Lightning Channels: https://lists.linuxfoundation.org/pipermail/lightning-dev/2023-February/003842.html

[TheBlueMatt]

I will be on a plane for this meeting, sadly.

[Roasbeef]

blinded paths:

• lnd doing interop testing now, using lnd to forward a blinded path route to CLN

onion messaging:

• starting to do interop testing for the flow of offers, but not the payment part yet

taproot:

• LDK now pushing towards interop testing
• laolu still needs to push out the lates script changes

attributable errors:

• formerly known as "fat" errors

• now the TLV portion merged in, joost posted some additional rationale w.r.t why TLV

  • sender doesn't need to reveal characteristic about what they can interpret
  • nodes can just now return what they want, maybe with some priority ordering (?)
  • in case there isn't enough space, then routing nodes can make a decision about what's most important to return

• TLV ftw!

• dust exposure:

  • now updated to latest commit: 7705356
  • still needs ppl to take a look at

splicing:

• working on re-working the protocol to use distinct messages
  • was gonna use the channel ID, now just use the txid of the splice transaction
  • now used as the main identifier, now use that TLV in commitment signed
  • do you ever change the channel ID? no...a bit too tricky, is a universal identifier
  • what about invoices (scid)?
    • changes, maybe can just use scid aliases to make it easier?
    • also have a grace period, 6 blocks before forgotten
• new issue discussed:
  • partial mstat balance
  • then space in
  • what to do with it?
  • just hand it to miners, so at most 1 sat -- give it over to fees

CLN version of peer back up protocol:

• messages: 7 and 9
• has a blob other peer can store, then another message to return the blob
• experimental option in upcoming release, odd message -- some other thing going on
• eclair has version for?
  • #881
  • might be in chan reest (their version)
  • CLN version sends it before re-establish
    • is this counter to the spec?
    • no, since not specific to a given channel, just a blob of data, sort of just on the gossip later
• questions:
  • blob size restricted?
    • single message so 65k limited, srs node should have srs back up
  • how's it find the nodes?
    • got to the gossip network, potentially rediscover them
    • only private channels, would have more difficulty (especially if changed IP addrs)
    • something something sats backup, 65 KB storage slot

high availability channels:

• to prove that peers are serious
• can establish a view of the network totally, anyone can advertise
• if it's a free signal, then not costly at all, so why wouldn't everyone just set it?
• idealized to be a property of a direction instead of a channel

channel jamming:

• have been running meetings every other week
• carla has been taking the new chaincode paper to try to turn it into a spec/implementation
• early unconditional push fees along route isn't going to work
  • as you start to accumulate fees if you have high/low fees then ppl can just start to take the fee
  • if upfront fee > normal fee, then node would just always take
• proof of forward:
  • now trying to go back into approach where need to prove that you can settle the preimage
  • if you try to route and have two attackers, then can just take (can you if PTLC?)
  • also leaks forwarding information now, since if you fail back, then you never got it down far enough
• ppl now working on the endorsement and the unconditional push independently
  • may be trying endorsement with experimental feature bit, possibly run in read only mode