Mark “Murch” Erhardt and Mike Schmidt are joined by Lisa Neigut and Bastien Teinturier to discuss Newsletter #279.
The Bitcoin Optech Podcast and transcription content is licensed Creative Commons CC BY-SA 2.0
Selected Q&A from Bitcoin Stack Exchange
Releases and release candidates
Notable code and documentation changes
Mike Schmidt: Welcome everyone to Bitcoin Optech Newsletter #279 Recap on Twitter Spaces. Today, we’re going to be talking about updates to the liquidity ads specification; we have six questions from the Bitcoin Stack Exchange that we highlighted; and then we have our normal Releases and release candidates and Notable code changes to go over. I’m Mike Schmidt, I’m a contributor at Bitcoin Optech and Executive Director at Brink, funding Bitcoin open-source developers. Murch?
Mark Erhardt: Hi, I’m Murch and I’m back from vacation.
Mike Schmidt: We have two LN experts in the space this week. Lisa.
Lisa Neigut: Hey, I’m Lisa, also known as niftynei, a longtime contributor to the Core Lightning (CLN) project and author of the original liquidity ad specification.
Mike Schmidt: Bastien?
Bastien Teinturier: Hi, I’m Bastien, I’ve been working at ACINQ on the LN specification and Eclair, and implementing and thinking about liquidity ads for a while as well.
Update to the liquidity ads specification
Mike Schmidt: We’ll jump into the newsletter and talk about some of the liquidity ad specification discussion. Liquidity ads are a set of proposed changes, I think initially in the middle of 2021, to the BOLT specs that allow peers to signal fees that they would charge for committing liquidity to a dual-funded channel. Lisa, how would you add to that definition; and then maybe also explain the use case a bit for the audience of why we would want such a thing?
Lisa Neigut: Yeah, I think that’s a great way to summarize it. I think the reason why you would want something like this, so one of the general problems that an LN Node has when it joins the network is, that if it wants to receive payments, it needs to convince another node on the network to take Bitcoin that it owns and set it into an account for its own personal use. So, you sort of need to figure out one who has bitcoin that they could put in an account that I could use so that I could then receive payments over LN. You need to figure out who’s able to do that, and then typically the way that this has worked up until today would be kind of informally through friend networks, or you’d hop on to a website and make like rings of fire, I think they called them, to all open a channel to each other so you all can receive a payment from each other.
Or the general idea was like, okay, what if this has the looks of a market problem? Some people need to receive payments, other people have bitcoins that they want to make available for others so that they can receive payments. What if we use the existing LN protocol to build a decentralized bulletin board system that would let people negotiate in a P2P manner; so, there’s no central marketplace, all the negotiations about the buys and sells of inbound liquidity could happen directly between two LN nodes, so what would we need to make that happen? What kind of signaling would we want to have so you would know who to go talk to? So, it really takes some ability that we added with the dual-funded channels and makes it such that you can now use that to actually sell or buy liquidity if you’re a node. That was a long explanation. Anyways…
Mike Schmidt: No, that’s great. And this is specific to dual-funding, so this isn’t a marketplace for people to open channels to me, but to join in on a dual-funded channel.
Lisa Neigut: Yeah, exactly. And so, the way I like characterizing it or how these two things kind of build on each other is, dual-funding made it possible that, you know, channels are two sided objects, you’ve two parties on one side; the v1 of LN channel opens, only one side had an opportunity really to put funds in the channel at start, and then the other side would just be like, “Okay, cool, you’re going to open up a channel and I’ll be able to receive your money, that’s great”. Dual-funding changed that, so the v2 of LN opens lets both sides put money into a channel at start, so that’s new and exciting. But then you run into a problem if you’re on the side of someone who’s like, “Hey, I want to open a channel with you”. It’s deciding whether or not when that other person opens a channel, whether or not you want to put money in it.
So, liquidity ads, the way I think about it, also a coordination kind of mechanism, right? It’s a way that you can coordinate like, “Hey, I want to open a channel, you should also put some funds in it”. And the way that you coordinate is like, “Hey, I’m opening a channel, I’ll pay you some money to put some funds in it”. So, yeah, they go together, they build on top of each other. You can’t really do liquidity ads without that dual-funding, that ability to basically negotiate a bitcoin transaction with two parties, because we use that as basically our contract for opening it, and you get a signature from both sides on that opening transaction, and that kind of serves as your signatures on the contract of who’s paying how much for the inbound. It’s cool because you can see that the other party has actually put the money they said that they would into the channel open contracts. If you think of it as a funding transaction now as a contract between two parties that becomes real and live as soon as you get both parties’ signature on it, all of a sudden you can set up cool, new relationships where both parties are able to bring money to the table, and you can basically negotiate the terms of that new opening funding transaction because we’re using the v2 opens, which we’ve been calling “dual-funding”.
Mike Schmidt: And so, some of this discussion or some of the negotiation for doing this previously, you mentioned, was just outside of the LN, whether that’s me signal messaging somebody and encouraging them and providing my reputation in some way. But the proposal is to move that mechanism to being gossiped within the LN, you’re providing some information there, which is basically pricing-related or fee-related information, but it’s still on the node operator to vet any non-fee-related things, like reputation of the node or things like that; is that right?
Lisa Neigut: Exactly, yeah. So, you’re still operating in a marketplace, there might be people that buying from wouldn’t be very beneficial to you, maybe they don’t have any other channels, and so buying their liquidity inbound would be great only if you want to get a payment from them. So, yeah, I think there’s definitely other things you’d want to consider when you’re looking at all the candidates, the people that you can buy inbound liquidity from, for sure. The protocol itself, we don’t really deal with that in a protocol. The protocol is mostly about, “Okay, I’ve already decided that I want to make this trade, so who can I make the trade to?” So, it helps you identify people to consider with other heuristics, like how much inbound liquidity they have or other channels they’ve opened or how long they’ve been on, etc. So, it provides you with a set to start with that extra filtering around who your partners are. And then once you’ve decided who that you think you would like to at least attempt to open a channel, maybe go into the negotiations with them, so to speak, it’s how does that negotiation happen.
Mike Schmidt: And you’ve had a PR open for some time, and I believe CLN has implemented the original proposal. What’s the impetus for your mailing list posts and these changes to the proposal?
Lisa Neigut: Yeah, great question. So, our initial draft went out, I want to say in 2021. It’s been up on CLN. I think LN-Router has the marketplace where you can see all the CLN nodes that are advertising, and I think there’s about like 12 to 15 nodes on any given day that are basically participating in that market. So, the impetus of the change from the first version is first, there’s a couple of things that I put in the first version that I’m unhappy with, so I wanted to make some changes in that place. The second reason that I’m spending more time on it now than I have in the last few years is, we’re getting really close to getting that initial v2 channel open dual-funding protocol into the spec, so that really provides us with an opportunity like, “Okay, dual-funding is here”. We’ve spent a lot of time, especially Bastien in particular has been a huge champion of this v2 opens, the dual-funding stuff, and splicing, etc.
So, because of all the time, energy, and effort that he and his team have dedicated to that proposal, we’re basically now at the point where dual-funding is done, that’s almost in the bag, we’re really hoping we’ll have that merged into specs soon. That means that all these new proposals being built on that original spec proposal, like splicing builds on it, the liquidity ad proposal builds on it, it’s like okay, let’s get back to these other proposals now that the foundational layer that we needed is looking like it’s really solid. Okay, what’s next? Let’s get liquidity ads updated to a point where I’m happy with it, Bastien’s looked at it and is happy with it, the larger LN community has looked at it and weighed in on what they think are important things to consider or add or how to change how that protocol works. So, yeah, the impetus is basically dual-funding’s almost here and almost done, and we’re ready to move on to the next things that that unlocks for us.
Mike Schmidt: Go ahead, Bastien.
Bastien Teinturier: And another impetus is the high fees because one of the great, great benefits of liquidity ads for one example, one scenario for mobile users, is that they can buy inbound liquidity in advance when the fees are not too high so that when the fees get high like they do today, they already have inbound liquidity and don’t have to make any onchain transactions to be able to receive more funds, stack or send out. And since many users have been beaten by that, trying to stack money and having to make some spliced transactions all the time in a high-fee environment, they really want the opportunity to be able to prepare ahead and buy liquidity when o-chain fees are cheap, so that they don’t waste too much money on onchain fees later.
Mike Schmidt: That makes sense. And am I correct in understanding that in addition to your work, Bastien, on the dual funding proposal, that you and the team at Eclair are also looking at integrating liquidity ads into ACINQ?
Bastien Teinturier: Yeah, actually I had a draft PR for liquidity ads and a draft branch in Eclair that has been up for, I’m not sure how long, but at least a year, maybe two. And I’ve updated it recently to match the latest spec. And I now have a branch both on the Eclair side and on the Phoenix side to have both the seller and buyer code, which is not fully spec-compliant yet, because we were still discussing a lot of details with Lisa on things that should maybe change in the spec. But that’s a good opportunity to experiment, to prototype, to have real users play with the feature and see what works and what doesn’t, to make sure that we come up with a spec that works for as many scenarios as we can think of.
Mike Schmidt: Maybe we can get into it a bit. So, what does need to be changed? Lisa, you mentioned you had your own opinions on things you didn’t like about your initial draft. We talked about dual-funding getting close; okay, so that was sort of the impetus. We talked about the high fees as another motivation. So, maybe pick a few of the most impactful potential changes to the spec that are being discussed.
Lisa Neigut: Yeah, so I think the one that Bastian and I are still trying to figure out where we land on is this idea of having a lease. So right now, the way that the spec is written – if you have a lease, how does that get enforced in the contracts, and the contracts being these bitcoin transactions that you have about a channel, right? So, the current proposal, we’re like, okay, every single lease that you buy has a fixed duration. That duration is 4,032 blocks; that’s roughly about a month. So, the way that we’ve enforced that duration of the lease is that on the seller’s money in the channel, we put basically what amounts to a timelock on it, and if that channel closes before the lock is over, then the seller’s funds aren’t really – basically that liquidity becomes locked up and unusable to them.
We do this as an incentive for them to keep the channel open as much as possible and keep that liquidity available to the person who’s bought it, because if they keep it up and available, there’s always a possibility they could earn routing fees on that liquidity. They go to chain, it’s just going to sit there until the end of the lease anyway, so there’s no incentive for a seller to sell you, let’s say, a million-sat channel, and then ten blocks later, decide to splice it out of the channel into a new channel where someone is paying them for a million sats. The point of the duration – I think, yeah, Bastien’s raising his hand. He has, I think, some really good points though about some of the downsides of locking your funds up into a duration. So, I’m going to kick it over to Bastien to chime in here.
Bastien Teinturier: Yeah, one issue for example with that is that it looks like it works, but actually you can enter into a cat-and-mouse game where each side tries to blackmail the other, because for example you buy liquidity from me for a month so that I cannot take that liquidity out. So, if I want to splice it out, you should just reject that splice-out attempt. But if you want to splice-in some funds and take the opportunity of your splice-in to try to add outputs on my side to splice things out, then either you accept it and I just took funds out disregarding the lease, or you reject it but then I’m preventing you from splicing-in funds into the channel. So, both sides can harm the other side and try to work around basically the lease duration.
The main issue I have with the lease duration is that it protects the buyer, but it creates issues on the seller side. Because if, for example, you have a channel where the buyer of liquidity initially has all the money on their side, for example 1 bitcoin, and wants to buy a very small amount of inbound liquidity for one month, they’re going to pay a very small fee because they’re buying a small amount. But then using LN, they push all of the liquidity on the other side of the channel, which means that the seller has moved some of their outbound liquidity in other channels to their own counterparties. So, now the seller has a large amount of liquidity in that channel that is locked for a whole month, but the buyer only paid for a very small amount of money. So, then you can enter into a blackmail game where the buyer can say, “Okay, I’m going to let you get your funds back, but only if you pay me a fee”. And I think that can be exploited to grief all the sellers, and I think this is an issue, so I’m not sure yet what we should do.
I think that what we discussed recently with Lisa is to make that lease duration optional, so some people, if as a seller you are afraid of being griefed by that, you don’t offer lease duration at all, you don’t really lock your output and the buyer only relies on incentives from you not to cheat. And if you are not a reputable seller, then you’re going to offer that guarantee to buyers, but you are taking a risk that the buyers lock your liquidity, lock more liquidity than they have paid for, for a potentially long duration. So, there’s a trade-off here that is hard to navigate. And I plan on making a mailing link post to see how people think about this issue and if people have ideas on how we could work around this. But I think this is really hard, because protecting the buyer means giving them a griefing opportunity on the seller.
Mark Erhardt: Would it work to – sorry, you go ahead.
Lisa Neigut: Yeah, I was going to say, this is where Bastien and I were talking earlier is, I was trying to make the point that this is why I think a marketplace is a good thing and having different options for different cases is maybe the best we can do as a protocol designer, where sellers that are worried about the duration risk, being they get their money locked up and they’re unable to reuse it otherwise, one, either they should price it – well, splicing I think added some new wrinkles to this which didn’t really exist before, because you used to have to completely close the channel. Splicing offers new opportunities which I think the interaction with liquidity ads is something that we’re still working through. So, I think that’s definitely a really important thing that Bastien has raised for us to consider.
But one thing about it is, really when you’re pricing – so, the original proposal for liquidity ads, we were asking you, we were saying, “Hey, there’s one duration, it’s about a month of this contract. So, when you’re pricing it, think in terms of what you want as a node who’s selling that liquidity, what would it take for you to lock up funds, guarantee that these funds might be locked up for a month, no guarantee they get used, what would you price that at?” Now we’re saying, “Hey, this duration is variable, so your node probably needs to have a way of repricing the liquidity based on the duration that the person who’s buying it requests”. Maybe they request a zero lease, which is what Bastien thinks is maybe the best for big sellers who are worried about a node that wants to just lock their liquidity up into these contracts, and then they can’t really move it around because they get stuck. Suddenly you end up in this kind of interesting price curve, almost like the US Treasuries, so to speak, right? You know, you lock your money up in a 30-year, you’re supposed to get paid more than if you lock it up in a 10-year, etc.
So, yeah, this question of duration is, again, from the seller perspective, it’s a risk on their side that their funds might actually get locked for that duration. And as Bastien is explaining, now with splicing, there’s an opportunity that maybe a malicious buyer on the other side, channel counterparty, would decide that for whatever reason, they really wanted to make as much of your liquidity as possible locked into this contract. So, they do some things with adding more funds than the original lease and use the onchain stuff that we’ve got in the protocol to keep the funds from being able to be moved elsewhere. Yeah, but anyways, so these are the things that we’re kind of rethinking through. And that big one, I think the biggest difference one is, the duration in the original proposal was fixed and known, and this new proposal we’re like, “Okay, do we even need a duration?” If we need a duration, maybe we should make it a flexible variable number.
This makes the marketplace a little harder to figure out, because now all of a sudden you would assume different rates for different durations. The old protocol, it was a fixed rate, everyone advertised rates for the same duration, the market was uniform. Now it’s like, okay, duration is up in the air, maybe it’s negotiable, maybe there’s no duration. And so we’re trying to figure out, okay, how do we help people figure out how to price that; what kind of protocol do we need? Suddenly there’s like a little bit of variability in it, that sort of thing.
Mike Schmidt: We have a speaker request from Chris. Chris, you have a question or comment?
Chris Guida: Yeah, hey guys. I was basically just going to say what Lisa said. Basically, my thinking was that the initial lease fee would basically contain all the information as to the risk to the seller. So, if they have to put their liquidity in for three months, presumably that would be a much higher initial lease fee than just a week or something.
Bastien Teinturier: Yeah, but the issue with that is that the seller does not know how much capacity is going to end up being locked. Because you sell, the buyer buys from you three months of 10,000 sats and you take a proportional fee on those 10,000 sats. But then by splicing-in and moving funds to your side using LN, you end up locking, for example, 1 bitcoin or 2 bitcoin for three months, and you cannot price that in beforehand because you –
Mike Schmidt: T-bast, I think you’ve cut out.
Mark Erhardt: Oh, I thought it was my desktop here.
Lisa Neigut: No, so really I think what Bastien is focusing on, and I think it’s worth thinking about, really has to do I think with, you know, LN liquidity is liquid, which is kind of cool. I could have a balance of, let’s say, 5 million bitcoin in a channel, my peer has a whole bitcoin, they’ve bought that 5 million, right? But then I have on my other channels more balances that I’m able to push out to the other side. I think what Bastien’s worried about is that your channel peer is going to buy some liquidity from you at a low rate and then push their side of the channel over to you, which then pushes out the liquidity all of your other channels. So, I think you are going to get paid for that movement of funds, whatever you’re charging in a channel fee to move it.
But the downside then is that until or unless those payments get pushed back to the other side of the channel, now you have bitcoin which maybe in those other channels was unencumbered and didn’t have a duration lock on it. Now you’ve basically exchanged that unencumbered bitcoin for this encumbered bitcoin that’s under this lease duration and basically I think the point is, there was no point at which you were compensated minus those channel fees, which presumably you’re charging at whatever the routing market rate is, and doesn’t include maybe an extra benefit of now basically moving those funds from unencumbered channels into holding that balance in an encumbered UTXO, which I think is valid.
Some of the things we’ve considered is, maybe there’s a way we could change the way that the duration is held in that channel contract with the original buyer of the liquidity lease, such that now there’s two UTXOs, one of them is encumbered, an encumbered one is never more than the amount of the lease. But then, Bastien was making some, I think, very, very good valid points that when you start adding that kind of construction, you end up with questions of like, “Well, we offered this Hash Time Locked Contract (HTLC) to our peer, but it got bailed back. Which of these two buckets do we put it into?” etc. So, yeah, I think we’ve still got, in terms of protocol design space, a few more things to figure out exactly around this duration question. So, I think, yeah, this is a great forum to bring these questions up. So, Mike, thanks for having us on and, Bastien, thanks for joining and bringing up all these great issues.
Then, I think the mailing list, hopefully I think Bastien said he’s going to do a good writeup of this for the mailing list so we can get it out there. I don’t know if Bastien’s back or not. Maybe he can talk us through, like I think Bastien’s proposal was we just get rid of duration and so you no longer pay for duration. The downside for that is if you’re a buyer, you no longer really have an option to get a guarantee that you pay one basis point for a million sats in your channel, and then now with splicing, someone could splice that money out immediately, the next block basically, to another channel, where someone’s now paying them another basis point for those funds, so you had access to them for maybe a block before it got reallocated away. Again, this is kind of where I think one of the nice things about it being an open marketplace and different nodes, and being able to have a pricing curve around duration is, maybe as a buyer you would see ACINQ’s offering a zero-duration lease for a cheaper price than maybe another node that you don’t know as well, who will give you a thousand-block duration on that on that liquidity as a guarantee, but they’re going to charge you a little more for it.
Mark Erhardt: Could you clarify something? You only pay a lease on the amount that the counterparty provides, right? There’s no contribution to the lease price based on the total capacity of the channel; is that right?
Lisa Neigut: That’s right, yeah. So currently, the way that the protocol is organized is that you, as the buyer, pay for the counterparty’s funds that they initially contribute, or at the time that you’re – we’re talking about adding this protocol to splicing, so you’d be able to use a splice to ask your peer to basically put more funds in the channel. So, at whatever point you’re doing this liquidity, it would be for whatever capital the person that’s selling it is adding to the channel; that’s right. Another thing about the protocol that I think – oh, go ahead.
Mark Erhardt: So, follow-up question maybe. Doesn’t that mean that the original liquidity ad is sort of mispriced because it doesn’t cater to how much the buyer wants to add? And so, what t-bast said with, if you splice-in more funds you might have a higher risk as the liquidity ad provider because now more funds on your side could be locked up for the duration, but don’t you already have that based on who might take your offer in the market in the first place, because if someone with a small amount takes it or someone with a huge amount takes it, you have completely different outcomes?
Lisa Neigut: I think, yes, I think that’s true. I think the general understanding of the protocol has been that the reason someone is paying you for that in the first place is that they want to receive it on their side. So, I would expect in general, most liquidity ad buys would be fairly one-sided, but I might be wrong about that. Maybe you’re paying someone so you can get a balance channel, so the channel opens with balance and you’ve paid them for their half of the liquidity. So in that case, yeah, there’s this risk, so to speak, that 2X of your original capital, depending on how the flows go, might end up basically encumbered under the duration.
Mark Erhardt: Okay.
Lisa Neigut: Bastien had his hand up.
Mark Erhardt: Oh yeah, sorry, I can’t see that!
Bastien Teinturier: Yeah, so I think that the really interesting thing is that, as Lisa said, since balance in LN channels is liquid, what you buy when you buy a liquidity ad is quite subtle, basically. If there’s no lease duration, no CHECKLOCKTIMEVERIFY (CLTV) that are being added, then you only buy the fact that the other side, at this exact moment, is going to be adding that amount to the channel, and then you don’t know what happens to that amount later. If you add a CLTV, what you are buying is that the seller is going to add right now some amount, and then for the duration of a channel, they won’t be able to take any of their balance out of a channel. But that balance could be anything between the amount that they’re adding now and the whole channel capacity, and the channel capacity in itself is something that can change due to splices. So, it’s a bit fuzzy what exactly you are buying.
From the buyer side, the case where there’s a CLTV is really great because you are actually buying more than what you think you are buying. But from the seller side, it’s a bit dangerous because you don’t know exactly how much you’re selling. Does that make sense?
Mark Erhardt: Yes.
Lisa Neigut: Yeah, it totally makes sense. Oh, I’m sorry. I think, yeah, I think Murch brings up an interesting point. Maybe there’s a way we could add a residual cost or an additional pricing on the total balance of the channel, so there’s some way that the total balance gets kind of priced into the liquidity buy. So, it’s not just the funds that the person’s putting up, but some risk factor of the total balance maybe being put under that duration that would be added as a cost to the buyer, is a thought of one way of approaching this particular risk. So, that would mean that the seller of liquidity would be compensated for the additional risk of a total balance of funds being basically committed to that duration.
Mark Erhardt: Yes, and if you have that concept already, you could also reprice the lease at the moment that someone wants to splice-in more funds, because now they’re changing the capacity and the capacity is already part of the pricing model. But yeah, you get a multidimensional lease cost now.
Bastien Teinturier: Yeah, exactly. It’s really hard to constantly repay fees whenever a splice happens, because there are even cases where the splice makes it so that you don’t even have enough funds to keep paying the fee. Or it creates an explosion of cases where I’m pretty sure there’s going to be edge cases that just don’t work if we start doing that, because repricing and repaying a fee every time you do a splice sounds a bit dangerous. Especially, for example, if you’re splicing-out, then should you get back some of the fees that you paid before?
Lisa Neigut: Yeah, these are all great questions about contract setting. I think this is why the original version of liquidity ads, we had one duration and splicing didn’t exist. And so it was like, okay, that contract is a little easier to negotiate. It still has this risk I think, that Bastien pointed out, of the funds on the other side of the channel moving into it. But overall, it was a little bit of a simpler landscape to navigate. I think now, we’ve definitely got a lot more – we’re trying to make the protocol as flexible and be as useful as possible for as many use cases as possible. And there’s definitely, I think, some edge cases where we end up in places where, I wouldn’t call it undefined behavior, but you start running into cases where it’s like, okay, this isn’t simple to navigate anymore in terms of a contract, and I think that’s difficult.
I think this is one of the things that we’re working on as protocol designers and why we’re having these conversations, is because we do get to bring up things like, “Oh no, there’s all these edge cases, how do we how do we figure this out?” And then, yeah, the greater question is like, I think I think in general liquidity ads are a good idea. That good idea is like, “Okay, I have liquidity I want to sell and someone out there is probably willing to pay me for it”. The details that we’re getting, it’s going to take some time, I think, for us to figure out. Like Bastien was saying, being able to experiment and try out different things is probably a good place, and I think that’s really where we’re at with the protocol. But to figure out how do all these different – because now we’re really talking about, what are you buying; what’s that worth; what are you willing to pay for it? And those are all more economic questions that I think sometimes there aren’t really exactly straightforward answers to. Okay, I think Chris has his hand up.
Chris Guida: Hey, yeah, just one thing to clarify. When you guys talk about splicing from a channel lease, that’s after the original timelock expires, right, because if you tried to splice before the timelock expires, then that just wouldn’t be spendable, right?
Bastien Teinturier: No, you could, it depends. We can just choose that. But the issue is that if you are inside the lease and the seller’s main output has a CLTV, if a seller tries to splice-out, then obviously the buyer should reject the splice-out. But then the seller can also reject all the splicing and splice-outs that the buyer tries to do. So, do you want to allow something? And if you want to allow, for example, because it would make sense to allow the buyer to splice some funds out because the buyer has not sold liquidity, so the buyer should be able to splice-in and splice-out. But if we use dynamic fees depending on the channel capacity, what happens then when the buyer changes the capacity of a channel, especially if they reduce the capacity of a channel; should they get some of the fees they paid before back? It becomes hard, I’m not sure what to do here.
Lisa Neigut: Sometimes when I think things are getting complicated, I find it useful to go back to the original design goal, right? So, I think Bastien said it really well at the beginning, what you want is you want people who want to receive payments over LN, give them an opportunity to secure that capacity to receive a payment in LN ahead of maybe when they need it. So, maybe there’s a low-fee environment, you want to know maybe – I tend to think from the vendor use case on LN, like I’ve got a little shop, or I’m an Amazon, and I’m going to be making sales over LN; I am going to have a constant need for inbound liquidity, basically the ability to receive payments over LN. And it’s worthwhile for me to basically pay upfront to ensure that ability to get payments.
So, as a protocol, what we want to do is we want to enable people who know that they’re going to need, or have an expectation of receiving payments over LN, the ability to basically for a price, get that ability, have that ability and have that assurance and they’re willing to pay for it. So ideally, that inbound, you start looking at it, it’s like, okay, well, they’re only paying for so many sats, and generally they probably want an assurance that they’re going to be able to receive that much in LN for a time period. As a buyer, you’re probably going to want an assurance that you’re able to receive a million stats over like the next week, and that’s what that’s what you’re paying for. So when you start getting it, it’s like, okay, how can we design a protocol where that desire to be able to receive a million sats over an LN channel is something that we can guarantee? We’re not guaranteeing that they’re able to receive it and then send it back and then receive it again, it’s like a first time receipt. So, maybe there’s a way that we could redesign the contract such that they have that million stats available, I don’t know.
All these edge cases of like the money’s gotten pushed back and now it’s gotten bigger, etc, is really a result of the initial design that we came up with of how to, as much as possible, guarantee that that seller is going to leave that million sats available for them. Yeah, so maybe there are some other ways we could structure these in-channel, onchain, bitcoin transaction output contracts that are able to more nuancedly, I think, express that, because it’s not a design goal that they’re able to push back a bitcoin, and suddenly the seller has a bitcoin that’s been encumbered, right, that’s not something that we’re trying to design for.
Mike Schmidt: It sounds like there’s still some discussion that’s going on here for sure. So, if folks are listening or reading the transcript of this discussion, obviously if you feel like you want to contribute to that discussion, it sounds like that feedback would be valuable from the proposal authors. So, feel free to jump into either the PR, I guess, or the mailing list to provide your feedback. Lisa or Bastien, any other calls to action or parting words?
Bastien Teinturier: No. Just stay tuned for more changes and more feedback on the mailing list.
Lisa Neigut: Yeah, I think Bastien said he’s going to, and correct me if I’m wrong, is going to have a new update summarizing a lot of his discussion and some of the downsides he’s come up with, with the most recent protocol design that I’ve put out. So, I think that would probably be the best thing to respond to once that gets out, unless I’m wrong about that, Bastien.
Bastien Teinturier: Yeah, exactly.
Mike Schmidt: Well, thank you both for joining. You’re welcome to stay on to talk about the Bitcoin Stack Exchange questions and PRs this week, but if you’re busy, you’re free to drop as well, we understand. Next segment from the newsletter is selected Q&A from the Bitcoin Stack Exchange. I’m going to re-invite Murch since she left, so one second. Okay, Murch, you should have been invited. So, the Stack Exchange Q&A is a monthly segment that we do, where we go through the Bitcoin Stack Exchange and find interesting questions and answers and try to highlight them for the community here. We have six this month.
Is the Schnorr digital signature scheme a multisignature interactive scheme, and also not an aggregated non-interactive scheme?
The first one is, “Is the Schnorr digital signature scheme a multisignature interactive scheme, and also not an aggregated non-interactive scheme? So, there’s quite a few things going on here. Each of these could probably have its own Twitter Space about, but sipa, Pieter Wuille, described the differences between some of these terms, including what is key aggregation, what is signature aggregation, what is multisignature, and what is this thing in Bitcoin that we call multisig. And then he also, based on the definitions of those terms, jumped into a bunch of related schemes that are applicable in Bitcoin, including schnorr signatures as defined in BIP340, MuSig2, Frost, and I don’t think this one is actually applicable to Bitcoin, but the Bellare-Neven 2006 scheme. I don’t know, Murch, how much you want to dig into any of those, or if that’s an exercise for the listener. What do you think?
Mark Erhardt: Maybe more meta. We’ve had a very active new user on our site lately, and they have been extremely inquisitive, especially digging down into the details of a lot of questions. And Pieter’s been very generous with his time to answer a lot of their questions. So, I think the product of this, they’ve also added like 30 comments below where they continue to iterate on the topic. I think we’re getting a bunch of really comprehensive answers out of this. So, if you have ever been wondering about this, we had previously already a page on Optech to go into the details between some of these terms, especially in the context of multisignature versus multisig. But yeah, what we call schnorr sigs is not that well defined in Bitcoin. We actually use BIP340 signatures. And schnorr signatures, as defined, might be most clearly what was proposed in that paper some, what is it now, 30-odd years or 40-odd years ago.
Mike Schmidt: Yeah, Murch, you mentioned a great resource. If folks want to look into Pieter’s answer as maybe a first pass on Pieter’s explanation of these terms, and then we also do have a great multisignature page on the Optech Topics wiki, where you can kind of compare and contrast what is FROST and how does that fit in versus MuSig, etc, threshold signatures, etc.
Is it advisable to operate a release candidate full node on mainnet?
Next question from the Stack Exchange is, “Is it advisable to operate a release candidate full node on mainnet?” And I suspect, well I think we’re talking about Bitcoin Core, which makes a lot of sense because we do have Bitcoin Core RC2, and now 3, that we’ve covered in the Twitter Spaces. So, it sounds like this user is considering running that on mainnet. And I think the specific question was, is this a risk to the Bitcoin Network? Murch, you were one of the people that answered this question. So, what’s your take on whether it’s a risk to Bitcoin Network and what risks there may be, if it’s not, to the Bitcoin Network?
Mark Erhardt: Yeah, so I would generally not consider it a risk to the Bitcoin Network. The Bitcoin Network is composed of a number of nodes that operate under different rules. We’re supposed to be compatible with any version of Bitcoin Core node that’s been ever published. We’re also supposed to be – well, that’s not entirely true. The P2P message has changed in 0.3 or so, so any since then. But there’s also a bunch of other implementations of the Bitcoin protocol that we also have to be able to communicate with. So, whatever a malicious attacker could do to us by misbehaving inside or outside of the protocol is probably going to be much worse than whatever an almost finished, final candidate for release would be doing to us. So, the danger here is not that great. You should absolutely go ahead and run RCs in production, as long as you do not depend on the node operating continuously and well, because all of your business back end is attached to it.
So, I would say if you are a significant economic actor in the network, if you’re running your own wallet off of that Bitcoin Core node, if you, for example, have an indexer or other services attached to that Bitcoin node or maybe an LN node, then you might not want to switch to a release candidate. You might even want to hold off on switching to a new release for a month or two to let other more adventurous spirits test it further. But if you do run a full node just for the heck of it, or if you even run multiple full nodes to be abreast of recent changes, or if this is your testing setup because you do run business infrastructure running on a full node, you should absolutely run an RC and test it and report anything that is strange or unexpected about it.
What is the relation between nLockTime and nSequence?
Mike Schmidt: Next question was, “What is the relation between nLockTime and nSequence?” And really I just used this particular question as a jumping-off point for six related questions from the Stack Exchange this month. Murch, you mentioned that active user, LeaBit, I think it’s pronounced, who joined the Stack Exchange in the last month, and I think this person asked all six of these questions, really digging into timelocks. And it’s kind of encouraging to see somebody new on the Bitcoin Stack Exchange that is asking fairly intense technical questions. Oftentimes, as a person who authors a section of the newsletter, you can see that there’s new people joining and asking questions and sometimes they’re not as in-depth as this, so it’s encouraging. Hopefully, LeaBit can continue to level up their knowledge and maybe even start contributing at some point.
I don’t know, Murch, if you want to jump into each one of these six questions or if it would make sense to just maybe summarize nLockTime and nSequence, and if people are curious about the nuances, they can jump into these answers.
Mark Erhardt: I think let’s go with a summarized approach, because six questions is going to take a lot of time. NSequence is a field on every input. It was originally intended as a way to make sure that transactions could iterate through multiple versions, and it was supposed to sort of demark inputs with an increasing sequence number, and a transaction with a higher sequence number was supposed to be preferred over a transaction with a lower sequence number. However, there was never a mechanism that could actually enforce that behavior. So, people would have always been incentivised to simply accept whatever transaction pays more fees in order to collect a higher total revenue in their block. So, that mechanism was removed at some point, or never really worked. And later, that field that was required by consensus rules got changed in its definition and is now used to generally signal finality of transactions.
So, if all inputs on a transaction have a max value sequence that signals that the transaction is final then the LockTime will be disregarded. If at least one input has a value that is lower than the maximum, the LockTime in the transaction header, so transaction-wide, will be respected. So, that is either a block height or an absolute time defined in chunks of 512 seconds times whatever you set in this value since the Unix time epoch. And yeah, so the LockTime is a transaction global value of a single field on a transaction. It can tell us that a transaction cannot be included in a block up to the height that the number defines, and similarly for median time past. So, if the LockTime were block 100, the first block that a transaction can be included in is block 101. And the sequence, as I said, can be used to either enforce or not enforce the LockTime. And then for even lower values, first we get replaceability. So, for max-minus-two or lower, a transaction is considered replaceable per the BIP125 rules. And then for even lower numbers, we get the behaviors for CHECKSEQUENCEVERIFY (CSV), which I don’t want to get into detail now unless you make me!
Mike Schmidt: I won’t make you. I’ll just comment maybe at a high level that the constraints within Bitcoin have spurred some creativity in this, and timelocks is one example of that, of using these scarce fields that may not be being used for certain purposes, because you can’t just add new fields willy-nilly in Bitcoin. So, the creativity is admirable, but it does add to some complexity when trying to explain what exactly these fields are. So, if you’re curious about some of the interplay between these fields, check out the newsletter where we link off to these different interactions between things like CLTV, BIP68, etc.
Mark Erhardt: Maybe one small addition still here. So, it’s pretty funny that since we have CSV both with the absolute time and a block height, and also CLTV with an absolute time and a block height, and I think I’m actually not 100% sure whether LockTime can only define a block height or a median time passed, but I think it should also be congruent in that, so you actually get an explosion of possible combinations, and you cannot use some of these together. So, for example, if you have an input that defines a relative number of blocks that a UTXO is locked, it cannot be used together with a relative time passed, because both of those have requirements on sequences. And I think the sequence system downstream causes a requirement for LockTimes.
So, we have a Bitcoin Core contributor that’s been looking into automatically distinguishing which inputs can be spent together that are using different types of LockTimes. And yeah, it’s pretty complicated and interesting!
Mike Schmidt: And doesn’t RBF come into play there as well, too, right?
Mark Erhardt: Yeah, but RBF is just, is this considered replaceable or not replaceable? And basically anything that is below max-minus-one is replaceable. So, if it has a single sequence that is below max-minus-one, it’s replaceable. So, yeah, the whole LockTime interactivity, where the absolute times with the relative times, and relative times defined in height, block height, or in time passed, and also absolute times and height or time, they have funky interactions where some of them cannot be used together.
Mike Schmidt: We have a question from the audience. Larry is asking, “Would that be a consensus violation or non-standard?”
Mark Erhardt: The main issue is, you cannot have – so, the absolute times are defined in the upper half of the range, and the block heights are defined in the lower half of the range, and you can’t have both of those at the same time. You can’t have a value that’s both in the lower half of the range or the higher half of the range.
What would happen if we provide to OP_CHECKMULTISIG more than threshold number (m) of signatures?
Mike Schmidt: Next question is, “What would happen if we provide to OP_CHECKMULTISIG more than threshold number (m) of signatures?” The person asking this question wonders, what would happen if you had, I think in the example it was a 3-of-6 multisig, but then during signing, you passed five signatures to OP_CHECKMULTISIG? Sipa answered that currently this would fail, but also noted that previously you could have had additional signatures passed to OP_CHECKMULTISIG due to a bug in the initial implementation of the OP_CHECKMULTISIG opcode. So originally, well I guess currently, the OP_CHECKMULTISIG opcode has a bug where n, of the public keys, and m, of the signatures, are popped from the stack, but that opcode also pops an extra unnecessary item off of the stack, which forces all the multisig transactions to have a dummy value at the beginning of the unlocking script.
So, because of that bug, you could have potentially provided in the past an extra signature in that extra dummy value slot, and it would have been consensus compliant. But in 2017, there was a soft fork to not allow arbitrary data in that dummy slot. That dummy slot has to have a value of zero, and that was done in a soft fork, BIP147. Anything to add there, Murch?
Mark Erhardt: Yeah, also leaving an element on the stack would also make the transaction non-standard. And I think if a zero is left on the stack, it’s actually going to fail validation. You only can have positive elements left on the stack for it to be valid. Sorry, so to clarify, if there were a positive value left on the stack in the end, I think that would be fine. If there’s a zero left on the stack, I think that would fail. So, if you had multiple signatures added, it might actually pass, but be non-standard because there’s multiple elements left. But this is just spur of the moment. I’m happy to be corrected.
What is (mempool) policy?
Mike Schmidt: You mentioned non-standard there, which is a good segue to the next question, which is, “What is (mempool) policy? And Antoine actually went and started his answer with defining both policy and standardness. And Murch, I’m curious, as an educator and somebody who thinks about terminology, including being the author of the transaction terminology BIP, which doesn’t cover these terms, maybe how would you define policy versus standardness?
Mark Erhardt: That is an excellent question. Well, I would say that they are somewhat used interchangeably. When we talk about a standard transaction, we’re talking about what a default node would accept into their mempool. Standardness generally is a more restrictive rule set than consensus validity. So, we will accept stuff in blocks that we would not accept into our mempool. And what we don’t accept into our mempool, we do not relay on the network. I think for the most part, policy is referring to mempool policy. And what gets into our mempool, we consider to be standard transactions. But I think more strictly, standard is a specific function in the Bitcoin Core code that checks some of the mempool policies, but there might be further mempool policies that also restrict what we accept into the mempool. So, policy might actually be a little more restrictive even than what standard transactions are. So, yeah, I’d have to stare at that a little more. I haven’t read that answer yet.
Mike Schmidt: How about two examples from the answer that Antoine gave? One, he gives an example saying, “A transaction is invalid by Bitcoin Core standardness rules if it has a fee lower than 1 sat/vbyte”. And the second example is, “Sending an unconfirmed transaction which has 25 or more unconfirmed ancestors violates Bitcoin Core policy rules”. Would you categorize those examples as he had in standardness versus policy?
Mark Erhardt: Yes, this is actually exactly an example of how the policy might be even more restrictive than just the standardness. So, standardness here is something that applies to a single transaction that we can determine by just looking at that transaction isolation, whereas the policy that is being failed here is something that applies to a set of transactions. Concretely here, if a transaction has already too many descendants, we might not accept another descendant/if there’s too many ancestors already, we might not accept another transaction from that set into our mempool.
What does Pay to Contract (P2C) mean?
Mike Schmidt: Next question from the Stack Exchange is, “ What does Pay to Contract (P2C) mean?” and Vojtěch describes the P2C protocol, and we link off to our Topic wiki on P2C, and he also linked to the original proposal, P2C being a way that a public key can be created that commits to the text of a contract that a sender and receiver both agree on, which then allows the sender to later prove that the payment committed to that specific text or contract. Murch, would you care to augment that or does that satisfy the answer?
Mark Erhardt: That is satisfying.
Can a non-segwit transaction be serialized in the segwit format?
Mike Schmidt: Okay. Last question from the Stack Exchange, “ Can a non-segwit transaction be serialized in the segwit format?” The short answer is that you could do this, but in BIP144 that specifies serialization formats for propagating transactions that commit to witness structures, that part of the BIP states, “If the witness is empty, the old serialization format must be used”. And I think that maybe on its own is less interesting for the audience, but I did think it was interesting that Pieter, in answering this, also pointed out some interesting history, that Bitcoin Core version 0.14 through 0.18 unintentionally allowed using the extended newer serialization format even for transactions that didn’t have a witness. That didn’t cause any issues in particular, but it was interesting that that was there for quite a few versions. I don’t know if you were familiar with that, Murch?
Mark Erhardt: I was not, but I also want to clarify the answer states, it would accept it in deserializing. So, if it got a transaction relayed that had serialized a non-segwit transaction with the segwit format, aka it added the witness stack even though there were no witness items, then it would have deserialized it and accepted it, but it would never create those.
Mike Schmidt: Yeah, it would accept the old, but when it sends it out to peers it would use the new again. So, no I guess attack vectors there per se. That wraps up the Stack Exchange.
Core Lightning 23.11
Moving on to Releases and release candidates. Core Lightning 23.11 is out. Some things that I found interesting, I know we’ve talked about some of these and we’ve had PR discussions about these over the last month or so, but I’ll run through a few things I found interesting from the notes. More powerful access control for the rune authorization mechanism; improvements to the check command in CLN, and the check RPC is a command that verifies another command without actually making any changes. So, it’s a way to, I guess, check that your call to a particular RPC is doing what you want it to do, or would do what you want it to do; the ability to verify the validity of existing emergency backups using the decode command; large/wumbo channels; changes to the dual-funding process; new features for CLN plugins; and, “A whole lot of cleanups to stay on track with small specification changes”.
But this is just what I picked out, so obviously check the release notes for more. Murch, I don’t know if you had anything to add?
Mark Erhardt: No, but I was wondering whether nifty has something to add.
Mike Schmidt: Oh, good point. Lisa, if you’re still with us, do you have any comments on this CLN release?
Lisa Neigut: Hey, I think the biggest, exciting thing for me is the dual-funding interop that we got working with Bastien’s Eclair project, so that’s really exciting for us. It’s still under experimental and we’ll probably keep it there until it gets merged into the spec, which hopefully will be fairly soon. And so, I would expect in the next release of CLN, dual-funding will be out of experimental and something that will be able to be widely available without having to do any special configuration on your node.
Chris Guida is also reminding me that there’s a really cool new feature that got added to this version of CLN that I’m really excited about building on top of, something called, I think it’s called Unified Invoices, which basically now there’s a feature in CLN that whenever you make an invoice, it’ll give you an onchain address as well. And then if a payment gets made to either the invoice over LN or onchain, it’ll go ahead and mark that invoice as paid. This is really, I think, a really great way to make onchain and LN payments basically through the same CLN node, which is great. I think the feature to look for for that is invoices-onchain-fallback. It looks like Bastien’s got some stuff to add, so I’m going to toss it to him.
Bastien Teinturier: Yeah, it’s just regarding dual-funding. There’s one comment I made on the PR that we didn’t think about adding a Type-Length-Value (TLV) for the required confirmed input to the RBF messages. And if we use the same TLV bit as in open and accept channel, this is an even TLV bit, so it would be backward-incompatible. So, it’s nice that you are still flagging dual-funding as experimental so that we can do this thing in the spec PR and in our implementation, before actually shipping it and saying that it’s not experimental anymore. So, if you have time to look at that comment on the spec PR, maybe we can discuss it during the next meeting.
Bitcoin Core 26.0rc3
Mike Schmidt: Excellent. Thank you both for chiming in. We also covered the Bitcoin Core 26.0rc. And, Murch, you linked me to the 26 branch, and it looks like there are a few commits since the last RC based on timing. There were some around memory usage, some around compatibility with some Microsoft tooling, and some changes around the CI. And you also noticed something with regards to the ARM architecture in particular as well. Anything else you think folks should be aware of in this latest RC?
Mark Erhardt: No, as I said, I was out of the loop for two weeks, but it looks like a bunch of issues were fixed. So, if you’ve been testing or if you want to test the current RC, just pull it down from the website. There’s a binary, there’s of course the source code to compile yourself and give it a spin, we also have a guide for testing. There’s a bunch of suggestions of what you could try out or starting points. Please feel free to further take it for a spin if you find any of the testing topics interesting. And as always, if you find any bugs, please raise an issue in the Bitcoin Core repository.
Rust Bitcoin #2213
Mike Schmidt: Two notable PRs this week. First one is Rust Bitcoin #2213. Murch, you authored the summary of this PR for the newsletter, so maybe how would you summarize that for the audience?
Mark Erhardt: Yeah, so it looks like Rust Bitcoin changed the worst-case input size estimate. So, when you serialize an ECDSA signature, the r and s values actually have a non-fixed size. The way the DER rules work, if you have an r value or an s value in the upper half of the range, you have to add an additional byte in the front, because I think that’s how they make sure that the signed integers don’t get interpreted as negative. But anyway, so the signatures for ECDSA are actually not a fixed length. And high-s signatures got made non-standard in BIP66, or some of the cases at least – I think I said that wrong actually, sorry. In the 0.10.3 and 0.11.1 releases, people started enforcing stricter rules on how signatures should be serialized in Bitcoin. Before that, it was a little more Wild West. That’s especially related to how ECDSA signatures were implemented in the then used libraries. And then after that, I think Bitcoin Core especially, and other Bitcoin projects, started using libsecp for that sort of stuff.
So, anyway, high-s signatures have been non-standard for a long time, so we can assume that any wallet that tries to use Rust Bitcoin to build a transaction will only use low-s signatures, and therefore we can assume that a signature only has 72 bytes in the worst case instead of 73 bytes, and we can slightly reduce our estimate for the maximum size of the transaction and therefore use less fees to achieve a specific feerate.
Mike Schmidt: Is the saving of that byte the reason that those high-s signatures are now nonstandard and not really used?
Mark Erhardt: No, the main reason is that this was a vector of transaction malleability. So, a signature, the r value cannot really be modified by third parties because that would just break the signature. However, the s value can be mirrored along the axis. Plus-s or minus-s are both a valid signature for the same transaction. This has most notably been called out in the context of Mt. Gox because it, for a legacy transaction, changes the txid. So supposedly, Mt. Gox lost a bunch of money by not noticing that they paid withdrawals multiple times, because they built the transaction then tracked only the txid, but not the payment itself. And when someone malleated the signatures by turning around the s value, the txid changed and they never saw the transaction confirm and then made the withdrawal again.
So, supposedly that’s how Mt. Gox lost a lot of their funds. There may have been other issues. Also, you can, I think, WizSec wrote up a huge thing about that. So, if you’re interested in the history of how Mt. Gox got emptied out, you can read that. But anyway, the main issue is that being able to malleate the s value in the signature changes the txid.
Mike Schmidt: Yeah, that’s a great bit of history there with Mt. Gox. Thanks for adding that. Last PR this week, BDK #1190. So, previously in BDK, it was easy to get a list of unspent outputs, but harder to get a list of spent outputs. So, this PR adds a new method, called list_output, for listing all spent and unspent outputs all in one place. So, I think pretty straightforward. Murch, anything to add before we wrap up? It’s good to have you back. It was also good having Dave, so I hope you enjoyed your time off, but now we’re going to bother you every week.
Mark Erhardt: Yeah, I have nothing to add. Thanks for having me.
Mike Schmidt: Thanks, Lisa and Bastien, for joining. Thanks for Chris for jumping on and asking a question and Larry for asking a question, and for everybody listening. Have a good week. Cheers.
Bastien Teinturier: Thanks everyone.
Mark Erhardt: Oh, Chris has raised his hand.
Chris Guida: Hey, yeah, I just have two really quick points. The first is I’m going to answer my own question from earlier when I talked about how it seems like it’s impossible to splice-out of a channel lease. My understanding was that the timelock goes on the funding transaction, but it actually goes on the commitment transaction. And so, since those don’t go onchain, it makes sense that you should be able to renegotiate those timelocks. Thanks, Lisa! The second thing I wanted to do is talk about a project I worked on when I was with Start9 about a year ago. If you guys want to test liquidity ads, you can use CLN on Start9, and that has a liquidity ads feature. It’s kind of buried in the experimental section, but I’m interested to see if people try that out.
Mike Schmidt: All right, cool. Yeah, thanks for bringing that up. And thanks for speaking on the Space today.
Chris Guida: Yeah, glad to be here, thanks.
Mike Schmidt: Have a good week, everyone.