Mark “Murch” Erhardt and Mike Schmidt are joined by Peter Todd and Henrik Skogstrøm to discuss Newsletter #265.
The Bitcoin Optech Podcast and transcription content is licensed Creative Commons CC BY-SA 2.0
Changes to services and client software
Releases and release candidates
Notable code and documentation changes
Mike Schmidt: Welcome everyone to Bitcoin Optech Newsletter #265 Recap on Twitter spaces. It’s Thursday, August 24, and we’ll be talking about peer storage and fraud proofs for our news today. We have our monthly segment on changes to services and client software, and we have a few releases and PRs to review as well. We have Peter Todd as a special guest to help us talk about fraud proofs; Henrik from LN Capital as a special guest to talk about the Scaling Lightning initiative and the Torq v1.0 release; and we also have Max here, who’s going to be joining shortly, to talk about the Wasabi release that we cover later in the newsletter. We’ll do introductions. Mike Schmidt, contributor at Optech and Executive Director at Brink, funding open-source Bitcoin developers. Murch?
Mark Erhardt: Hi, I’m Murch, I work on Bitcoin stuff at Chaincode Labs.
Mike Schmidt: Peter?
Peter Todd: I’m Peter Todd, I do consulting.
Mike Schmidt: Henrik?
Henrik Skogstrøm: Yeah, Henrik, CEO and founder of Torq. We build node management systems and Lightning operation systems.
Mike Schmidt: Well, thank you all for joining us. We’re going to go through the newsletter sequentially here, starting with the news.
Fraud proofs for outdated backup state
We have one item this week, which is titled in the newsletter as Fraud proofs for outdated backup state. And this was motivated by a post from Thomas Voegtlin, who was unable to join us this week, who posted to the Lightning-Dev mailing list an idea to make Lightning channels resumable from a wallet seed, and these resumable channels would use OP_CHECKSIGFROMSTACK. The motivation here is Lightning-related, but it sort of went into a more generalized theme of using peers for your backup data in this specific instance, potentially Lightning channel backup data, and the associated risks and potentially fraud proofs that can be done when a peer is holding your data and is potentially fraudulently representing that.
Peter, I know you reviewed an early iteration of what Thomas put together. Do you want to try to summarize that, and then we can get into some of the other ways it could potentially be used, or potential downsides?
Peter Todd: Yeah. So basically to summarize, with Lightning, of course, you have the issue that you need to go backup state, and that backup process, of course, a problem with it is if you go use the wrong backup, your counterpart, you could go lose funds basically. This was a fairly well-known issue with Lightning. And an obvious potential party to go and do a backup with is your counterparty on a Lightning channel, but you run into the challenge of, well, what happens if they go and give you the wrong backup?
Thomas’s idea, which I think is kind of clever, and I’ll point out how it can be kind of generalized, is to go use the non-existing OP_CHECKSIGFROMSTACK opcode, as well as OP_CAT, which again doesn’t exist right now in Bitcoin. But basically, the idea with OP_CHECKSIGFROMSTACK would be to go check a signature on the stack against a message on the stack as well. And in the more likely version of OP_CHECKSIGFROMSTACKVERIFY, basically if that signature doesn’t correspond to that message, you would go and steal the check; fairly straightforward there. And in the backups case, I’m probably going to butcher this detail subtly, but essentially the idea is you go and have basically a time associated with the message you’re checking. And what the counterparty is promising to do is that for a given time, they will only ever return one channel state, which is the channel state that you uploaded at roughly that time.
So, the point of OP_CHECKSIGFROMSTACK is, well, if you can go prove that there are two different valid signatures on two different backups for that same time interval, that’s fraud. And if you go and make an output that has coins attached to this that is spendable by proving fraud, you basically have a bounty that can be collected by this proof of fraud. And the idea behind that is that provides economic incentive to not commit fraud, because if you get detected, if your counterparty does in fact already have – that backup doesn’t actually need your service, you’re going to go and lose those funds.
The way you’d implement this in practice is basically every time that you would go online, regardless of whether or not you had a copy of the data that you’ve backed up, and essentially regardless of whether or not you need the backup service, you would ask for the most recent data, check this, and if your counterparty lies, then you go collect the fraud proof. So, that’s the general idea there. And Thomas pointed out in terms of Lightning, basically anything where you needed a backup of a small amount of data that had this property, where you want to know what’s the most recent one and someone was willing to put up money for it, this could potentially be used for this.
Mike Schmidt: So, the idea is I can use my peer to back up some channel state information and I could essentially periodically test them to see if they would give me the correct data. And then, if I do have that data and they give me a fraudulent version of that, I can then have a bounty against that. I guess the assumption there is they wouldn’t know if I actually did lose my channel data, and so they would be incentivized to always give me the correct information if I was testing them or if I actually needed the backup?
Peter Todd: Exactly, yeah. So, the key challenge with this is you just need to implement it in such a way that the counterparty can’t figure out whether or not you’ve lost the data. A good example where they might have an idea that maybe you did, was let’s suppose that every single day, you connect 12 times a day, because your phone’s on and automatically connected, but once in a while, suddenly you go wait a week to connect. Now, a potential reason why you did that might be because you’ve lost your phone, and it took you a few days to get a new one and restore from the seed. So, that could be the result of a loss, but by putting up a monetary bounty, it disincentivizes testing this and there’s just less reason for them to go and try this out.
I’ll point out this tends to be more relevant in cases where your backup counterparty is, for instance, your channel counterparty, and they’re the people who might be able to go and take money from you by you screwing up and using the wrong backup. So, that’s like your likely outcome. And I’ll point out with taproot, you could even arrange for the bounty to be within the channel itself, so it would just be another taproot branch that could be revealed to go collect those funds.
Mike Schmidt: Murch?
Mark Erhardt: Yeah, I had the impression that this kind of mixes a little bit two different aspects. So, in the proposal from Thomas, I think he was actually not proposing to ask for the backup every time, but just occasionally. Whereas, there’s also a proposal that’s partially implemented by Core Lightning (CLN) and ACINQ that uses some backup mechanism and does ask for it every single time, but it doesn’t have the fraud proofs. So I was wondering, Peter, since you’ve read more about this, do you happen to know, what would happen if you actually lost your backup and you connect, and you of course ask for your channel backup, or I assume that you would be the first one to ask, but what if the other party asks you first and you don’t have it because you lost yours? So, would they just disconnect; would they just use that to steal from you?
Peter Todd: Well, I mean this is where the fine details of influencing this get a bit complex. I think in a case where you’re talking about, say, a private channel, where there’s a more clear relationship of the client, if you will, to the server, it’s certainly easier to say, well only one side’s going to ask for the backup and not the other. You can also say, well, whoever connects to whoever first. And again, in the case of a phone, well, that’s only ever going to go one way, because a phone doesn’t have a static IP address, so the counterparty will never connect to the phone. So, it gets down to the details. But certainly I think in a case where you were in a position to go and ask for a copy of the backup whenever you connect, I don’t think there’s any argument other than the trivial amount of bandwidth consumed to not asking for it every time. I mean, I think that’s the safest option and you might as well go do it.
Mike Schmidt: So another Lightning-related Service Provider that stores some data, albeit I think different data, would be watchtowers. Peter, is watchtowers a use case here as well for these sorts of proofs?
Peter Todd: I mean, for that matter, I don’t see any reason why you couldn’t simultaneously use this kind of technique and watchtowers at the same time. Maybe someone knows of a reason why, but I can’t think of any reason. And, the security in these techniques in the case of Lightning is additive, so you might as well do both if it’s possible.
Mike Schmidt: And you mentioned the need, at least in the case of this new type of channel funding transaction, which would include an additional taproot condition, that OP_CHECKSIGFROMSTACK would be required and is currently not available in Bitcoin. I think, is it true that OP_CAT would also then be required as well to be added?
Peter Todd: Yeah, I forget off the top of my head exactly why OP_CAT was needed in his proposal; probably something to do with exactly how the script worked. But certainly, OP_CHECKSIGFROMSTACK, that’s definitely needed, and it would certainly be possible to add it in a soft fork. You’d probably do it as a verified version, but I don’t see any direct reason why you couldn’t add it. There might be other concerns around covenants and so on, because certainly you can, in some versions of this idea, you can implement covenants with this. But that kind of gets into the details of how you would use it.
Mike Schmidt: Henrik, I’m curious, as an enterprise Lightning provider, what are your thoughts on maybe the general idea of peer backups or this proposal specifically?
Henrik Skogstrøm: Yeah, it’s really interesting. One concern, I’m not too deep into this topic, but one concern I have is the speed of updates. So, if you do channel backups essentially for very busy nodes, that’s going to be very, very fast. Yeah, what’s your thoughts on that, Peter?
Peter Todd: Well, I mean the backup, there’s an amount of data a counterparty already has to go and store anyway, just in the normal operation of a Lightning channel. And by proportion, I mean some relatively small constants, right? So, adding this just adds some constant factor more of data you would store, which I think is acceptable. Certainly it comes down to the details of what you’re willing to provide, but there’s no really fundamental technical reason why you can’t do this.
Mark Erhardt: I’m afraid I missed the first two sentences; I couldn’t hear you for a bit there. Were you saying that you wouldn’t be doing it every time your channel updates or what was your –
Peter Todd: Oh, I was saying you could do it every single time your channel updates, and the amount of additional bandwidth and storage necessary is a constant factor, right? So, obviously I think it’s true that it does cost bandwidth and storage space, but this isn’t an excessive amount. It’s not like we’re changing the BitGo notation to something fundamentally different. So, I don’t see any direct challenge.
Henrik Skogstrøm: So, what about the economic incentives? So, if the guy backing up the data cheats, then he gets punishment; but what if there’s a risk then for the guy backing up that something goes wrong? How’s the economic model around that?
Peter Todd: Well, I mean I think when you look at it, basically the – oh, I guess one way to put it is, in any Lightning channel, there is a risk for the counter party if they screw up, that’s just how Lightning works. You get punishments if you go and publish the wrong states on the chain. So, I don’t think this is fundamentally different. I think, depending on the amount at risk, they may choose not to, obviously. But, certainly from the point of view – I mean, think of it this way. If I’m the Lightning Service Provider (LSP), and I offer the service on incoming channels to me, that might be an encouragement to say, “Hey, on your phone you should connect to me, because it’s a bit safer on your phone”.
I don’t know that all this necessarily holds true in a more “P2P” Lightning model, but certainly the model we’re talking about, you know, people using their phone to open Lightning channels where frequently the phone is off and it could be lost, I mean I think it makes sense. And as a Lightning node, if you have very high confidence in your disk storage solution and backups and so on, which I think you need to anyway, I don’t see too much additional risk from doing this.
Henrik Skogstrøm: Right, that’s a very good point, super-interesting. So, it’s sort of different use cases. For enterprise, with a massive amount of rapid updates, it might make less sense than in those sort of rare cases for consumer wallets or mobile wallets. But yeah, maybe I’m just missing a fundamental piece, but yeah, I mean it’s just I guess you pay the guy doing the backups and then if he does something wrong, you sort of have a way to almost cashback or punish him more then.
Peter Todd: I mean, a point I would make is the payments for the backups in the case of an LSP is the transaction fees that they earn from having the channel open at all. So, provided that the cost of reliable storage is low enough for them, I think this is a reasonable service to offer.
Henrik Skogstrøm: Yeah, makes total sense.
Mike Schmidt: You heard Murch mention earlier in this discussion about CLN’s, at the time, experimental peer storage backup feature. And we spoke about that a bit in Newsletter #238. So, if you’re curious about this topic, jump back to that for an additional peer storage technique. And as a result of this peer storage being mentioned in this news item, we’ve actually created this week a peer storage topic on the Bitcoin Optech Topic List, which is sort of like a wiki, so you can see some other references to peer storage that we’ve had through various other newsletters previously. Murch, Peter or Henrik, I’m going to bring in an audience question here, but any other points you’d like to make before we wrap up?
Peter Todd: Yeah, well I’ll I mean I’ll just point out, my actual contribution to that conversation was, of course, just doing the math on the non-punishment version of peer storage, where every time you connect, you simply check if they gave you the right backup, and if they didn’t, you close the channel. And I think when you look at the economic math, provided that the rate at which the backup is needed is low enough, purely by transaction fees on the Lightning channel, it probably doesn’t make sense for a counterparty to attempt to defraud in most cases. Now obviously, I think this is a much weaker argument than if you actually have a direct punishment mechanism, but basically I think what the map says is that peer storage is valuable even if there isn’t Thomas’s punishment mechanism.
Mike Schmidt: Henrik, Peter anything else before we wrap up this news item?
Peter Todd: Nope, I’ll get back to my conference.
Mike Schmidt: All right, Peter. Thanks for joining us, especially on last-minute notice; and Henrik, thanks for chiming in.
Peter Todd: Thanks for having me on.
Mike Schmidt: Cheers. Next section of the newsletter is Changes to services and client software. This is a monthly feature that we have where we find projects that are integrating interesting Bitcoin technology that we talk about on the Optech Newsletter, and we like to highlight those for folks as a pat on the back for those pieces of software, and also the opportunity for folks to check out what’s being worked on at a few layers above this Core software that we end up talking about.
Scaling Lightning call for feedback
A little bit different than software update is our first item this week, which is Scaling Lightning call for feedback. And luckily we have Henrik on, who I think is spearheading a lot of the work around here, so I will let him frame this up, since it’s his initiative. Henrik, do you want to introduce what is Scaling Lightning? What are you trying to do; what can people do to help out?
Henrik Skogstrøm: Yeah, thanks. First of all, I sort of spearhead and started the project, but I want to also pull attention to the fact that Max Edwards is working on this full-time and has built basically all of the code we have today on the project, as he deserves. I can see he’s actually joining here, so Max, if you want to join, just request speaker; I think you have a lot of value to add there. But the core thinking behind the project is essentially, right now, there are some unknowns in scalability of the different Lightning implementations, and also perhaps certain parts of the protocol, and also with different software out there. And I’m not saying that Lightning doesn’t scale, I’m saying that no software is perfect and we need to thoroughly test things and prepare for the day where inevitably, there’s going to be a massive amount of traffic, and those changes can happen very quickly.
So, the project is essentially compiled by a set of different simple components. So, if you are a protocol researcher or you are developing the Core clients, you use it in one way; if you’re developing a software for a company, you might use it in a slightly different way. And when you’re writing end-to-end tests, you have a slightly different use case again. And lastly, if you want to run a node on the signet, you have a slightly different use case as well. So, that’s why the entire structure of the Scaling Lightning project is meant to be able to support all these use cases without adding a ton of complexity and trying to cover everything in a really sort of wordy way. Right now, it’s composed of simple components; the goal is to have everything as very small, simple components.
Firstly, we’re using Kubernetes at the core of this. And the reason for that decision is that you can run Kubernetes on a desktop by just clicking a button in Docker Desktop, for example. So, it’s very easy to run there, and a lot of production systems run through Kubernetes setups. So, we want to have that setup as most usable from development into production systems, and also be able to simulate larger networks in research setups, and also be able to deploy either one or multiple nodes in a signet and have a lot of activity in that signet.
So yeah, when we started down the Kubernetes road, it’s also where we’re looking at the test software that’s out there. It’s largely built by cobbling together Docker containers in some way and writing some scripts to run either a fixed set of those nodes or a variable set, and to start some channels between them, so you have this base network. You also have Polar, which sort of does the same, uses TypeScript to construct these different networks and has a convenient interface to do this. But the thinking essentially is that this is what Kubernetes is built to do. It’s built to set up different configurations of separate components running separate software, different nodes, etc. So, we’re leveraging all the different features we have within Kubernetes and Helm and Helmfile, etc, to not have to build all that from scratch, and also having a very standard way of configuring things that people are used to if they’re working already with Kubernetes or Helm.
Then, the second part is that all of these things that you set up through the Helmfiles, etc, they are very easily replaceable, so you don’t have to wait for the Scaling Lightning Project and developer binder to update something for you. You can just pull out any component you want, and that, of course, is very important for different implementations and different versions of implementations like CLN, LND, different versions of those. And it also allows you to create your own images and just replace them. So for researchers, that’s even more relevant where they can test implementation with some crazy changes to them.
Then, you have a set of what I would call like convenience add-ons. So, right now in the default setup, we start a sidecar client alongside each of these nodes to make it super-simple to communicate between them or with them from the client or through a library. And the point here is not to cover all the functionality that you can have in the Lightning node, but rather to do the basics so that that’s super-simple to do. So for example, have a command that says, “Open a channel from Alice to Bob of this size”, and that’s all you need to write in the terminal. Also equivalently, when you’re writing end-to-end tests, you can import the library in a set of different languages, so TypeScript, for example, and say like, “Give them the connection details, and open the channel from node A to B”. So that’s the sidecar client and the library helping you to do that very quickly.
Lastly, of course, get the connection details to each of these nodes directly, etc. So you can do whatever advanced setup you want and just leverage the fact that Scaling Lightning helps you spin up these networks of nodes for development or end-to-end tests. Yeah, maybe I’ve been speaking for a while. There’s more to talk through there, but I see you have a question, Murch.
Mark Erhardt: Yeah, so the way I understand you, you basically have a Kubernetes setup that allows you to spin up networks. So, for the LN, you can cover multiple different clients from different implementations, and you have an interface with which you can directly instruct specific nodes in that network to take actions. So that sounds like a great way to test interoperability and maybe the rollout of new features. Are you guys working on this by yourself or are you working with the implementations; is this a coordinated effort or is it just your project?
Henrik Skogstrøm: That’s a very good question. So first of all, I’m of course leading Torq and we’re building Torq, but this is completely separate from the company, this has nothing really to do with that. And the effort here is to create an industrywide tool that is used by companies’ implementations so that we pool our resources into making this as good as possible so that not each of the companies’ implementation have to write this software from scratch, each and every one. So, we’re talking with Lightning Labs, I’ve been talking specifically with Ryan Gentry; I’ve been talking with Rusty about this, Rusty Russell at Blockstream; briefly mentioned it to Spiral, but we probably need to speak more about it with them; also, Carla at Chaincode Labs.
This is something where we’re trying to build more support around and trying to build collaboration with different researchers and node implementations, but also companies. So, we have a Slack Telegram group where we have a lot of companies and people from the industry in there where we’re keeping them up to date, and we’re slowly now seeing more people engaged in trying it out. So yeah, long answer to that question!
Mark Erhardt: Yeah, okay, one follow-up question. So, I think there is a test network for the LN on signet; and does this network that you’re building, the scaling network, plug into the Lightning signet, or is it rather meant for everyone to spin up their own and to do their own research and testing, or does it all plug into one global landscape?
Henrik Skogstrøm: So again here, configurability comes in, but we’re not sure if we’re at that part yet, but the goal is to have one default one to encourage people to join an agreed-upon signet to get traffic there. But of course, for research, some types of research, you’re going to have some crazy shocks of activity, etc, so they can, of course, create their own if they need it, companies can create their own if they need that, they can use regtest locally, etc. But to just complete the thing on the signet, with the library, the second part of that is creating a random generator where we can create these sort of personas.
We can have a persona that simulates or works a bit, has a behavior of like Kraken and Bitfinex, these really, really large nodes. And then we have one that is more like a medium routing node, one that is like a pleb, a small node, and so on. So we can through that, for example, look at how the network looks today and take a subset. So, let’s say we have 15,000 nodes today, and then we take 150 nodes and we spin those up on the signet, on a cloud infrastructure, and give them different personas to make a mini version of the network, and then we use those random generators to create this activity. So also to do that, we’re thinking about opening up the sidecar clients in certain ways, so we constantly have this traffic and users or different developers, they have an active signet, which we don’t really have today. So, that’s one of the bigger goals of the project as well.
Mark Erhardt: That’s really cool. Last question from me for the moment, at least. I heard you mention that it also can be run on regtest, and I was curious about that. So regtest is, at least originally, construed to be the internal testing mechanism for Bitcoin Core. Why would you ever prefer a regtest over signet when you’re doing a network thing in the first place?
Henrik Skogstrøm: Yeah, actually internally now, in Torq we’ve been using signet with LND, for example. But the main thing is to have a very rapid development environment when you develop and when you do end-to-end tests. So, yeah, the main goal there is to have this development environment that is not testing in the wild-ish, right? So that’s the reason behind that.
Mark Erhardt: I see, thanks.
Henrik Skogstrøm: Great.
Torq v1.0 released
Mike Schmidt: Henrik, thank you for walking us through the Scaling Lightning initiative. We also, and this actually is coincidence, we also are highlighting the Torq v1.0 release, and that’s the next item from the newsletter as well. Do you want to talk about Torq generally and also some maybe high-level features in this v1.0 release?
Henrik Skogstrøm: Yeah, sure. And thank you for having me on board with this as well. I want to point out again that it’s completely separate projects. Scaling Lightning, it’s a community industry, and then we have Torq separate. Yeah, so v1.0 release was major for us. We’ve been live and people have been using Torq for a long time, but we got to the stage where we had full features in the basic management. We had multi-node management, we had automation coming and becoming very advanced, so we decided it was time to integrate to v1.0.
First, maybe I can just quickly explain what Torq is for those who don’t know. So, Torq is enterprise level software for managing Lightning operations. So, you can think about it as a node management system, but that’s at the beginning. So for example, the Lightning is a node management system for manually managing the basic actions of a node instead of using a command line. So, we have that functionality in Torq, but this is about more than that. So, we have very advanced automation, for example, where we’re replacing the need for creating most simple automation flows with the ability to do that with the UI and also do that through a workflow method. So, this is where you can basically make a decision engine. So, for example, “If your channel balance is lower than X, send me a message and close the channel”. Or, and then, for example, open a new channel to the same node. So, for example, if you always want a private channel to Cash App or River, then you can set that logic up so it always renews the channels when needed.
So yeah, that’s one part, and the v1.0 release there added some features, but it also added a visual logs functionality. So, when you run a workflow that closes channels, you want to make sure that you don’t mess up a filter. Let’s say you want to check that you open the channel always to a certain node, but only if conditions are right. If you mess up a filter there, you can end up closing all channels or opening a ton of channels. So, we built in a logging mechanism where you can both see what has happened on real runs, but you can also run dry runs. And then you can go in and see what actions were triggered, what was the input, what were the filters applied, and then basically walk through that visually. So, you have a diagram showing the graph of the decision engine. So, it’s actually like a directed acyclic graph, and you can click on the different elements and the different paths of that log and see just the relevant log items for that. That was a major feature in the Torq v1.0 release.
We also have in there a Hash Time Locked Contract (HTLC) firewall, which is similar to like CircuitBreaker, as that’s part of the workflow. And we have functionality to move funds between your nodes, which is another very important part of Torq, because large companies or companies generally run more than one node in order to have redundancy, and the ability to update one node while they’re still operational. So, multi-node management is also a major topic in Torq right now. So, I’ve been speaking for a long time there, maybe there’s some questions.
Mike Schmidt: None from me, Murch, you?
Mark Erhardt: No, it sounds good. I think I’ve heard so often in the last year that node management is a little bit of a headache and that running a Lightning node as an amateur is taking a lot of time. So, this sort of software sounds like it would definitely have a market fit somewhere.
Henrik Skogstrøm: Yeah, thanks. And to add there, Torq is very simple to use, so you can use it as a smaller node. But again, our mission is to reduce all the need for making the default software that every Lightning company needs to do. It doesn’t make sense if you’re building a wallet app or you’re building an exchange on Lightning, or something, that your customers don’t use your product because of your wonderful node and how well your node runs. They’re really using it and choosing your product for everything around it. So, our goal is to replace the need for all that basic functionality that you need in order for your company to get started on what’s actually important, and building the products that your customers love.
Mike Schmidt: Dawn, did you have a question, Dawn of an empire?
Dawn of an Empire: Sorry about that.
Mark Erhardt: I did not understand that.
Henrik Skogstrøm: Could you repeat the question?
Mike Schmidt: And he’s gone. Okay, well, Henrik, thank you for joining us for these first couple of items of the client services and software section. You’re welcome to stay on and chat the rest of the newsletter. If you have to drop, I understand.
Henrik Skogstrøm: Thanks for having me, I’ll hang around a bit.
Blixt Wallet v0.6.8 released
Mike Schmidt: Blixt wallet v0.6.8 released, which adds support for hold invoices and zero-conf channels. Maybe as a reminder of hold invoices, also called hodl invoices, instead of immediately locking in and settling the HTLC, when a payment arrives, the HTLC for a hold invoice is locked but not yet settled. And at that point, it’s not possible anymore for the sender to stop the payment or revoke the payment, but the receiver can still choose whether or not to settle or cancel the HTLC invoice. One use case of this would be, if there’s a merchant online selling something and taking payment, but only settling if and when they can actually fulfill on the item, maybe they need to go check the warehouse or some such thing, and once they see that they can fulfill, they can settle the invoice; and if not, then they can cancel the invoice.
Then for zero-conf channels, the PR feature noted that it isn’t zero-conf channels just for anybody, but you can add pubkeys for nodes which you would like to accept zero-conf channels from. So, that’s Blixt Wallet. Murch, any questions or comments there?
Mark Erhardt: None from me.
Sparrow 1.7.8 released
Mike Schmidt: Next release that we highlighted was the Sparrow 1.7.8 release, which added support for BIP322 message signing and also had improvements for RBF and CPFP fee bumping features. And the message signing for BIP322 is for single-sig addresses, but it does also include P2TR. So, that would allow a wallet to sign some sort of an arbitrary text string by producing a signature, and that’s done using a virtual bitcoin transaction. Murch, I think BIP322 is still in draft status. Are you aware of any progress around that BIP?
Mark Erhardt: Yeah, I’m afraid that it hasn’t gotten that much review, and I think it’s been basically just floating for two years. This may be the first implementation that I know of. I don’t think I’ve heard of another one before.
Open source ASIC miner bitaxeUltra prototype
Mike Schmidt: Next release that we highlighted here was the open-source ASIC miner bitaxeUltra prototype. And this is an open-source ASIC miner that is the third major revision of this bitaxe open-source mining chip. Essentially, they took I think the Bitmain 1366 ASIC and used that as inspiration, which was I think one of the S19 miners, and essentially this is an open-source project around that, so all the design files are provided. I thought this was a little bit interesting, even if outside the box of our normal topics. Murch, any thoughts there?
Mark Erhardt: First time I hear about it, to be honest.
FROST software Frostsnap announced
Mike Schmidt: Next piece of software that we highlighted this week was Frostsnap, which is an interesting team that’s put together a project based on FROST where, okay, so I guess maybe we should step back and talk about FROST. So, FROST is a threshold signature scheme where you can have k-of-n signatures, as opposed to something like MuSig where you have m-of-n signatures. And their goal is to, “Provide access to funds through a threshold number of devices”, and they have a few different features that would be available given this FROST architecture. So, “Users would be able to change devices belonging to their multisignature without having to make any onchain transactions. With Frostsnap devices, users would be able to add new signers to your key. If you lose a device, you can replace it. If a device is stolen, you can make every other device incompatible with the compromised ones”. And it says, “We can adjust our multisig as necessary while still controlling the same Bitcoin wallet and private keys”.
What’s underlying this Frostsnap software is this secp256kfun library, which is something that’s been in development for some time, based on GitHub, and that’s authored by Lloyd Fournier. Murch, did you get a chance to look into this at all?
Mark Erhardt: I did not, but I’m familiar with the ability of FROST setups to switch their keys under the hood, and the way that that works is in this threshold signature scheme. Of course, you only need signatures or participation of a subset of keys up to the threshold in order to provide a sufficient signature. So basically, what you do is you use the key material from that number of keys to generate unrelated new keys of a new quorum, and then this new quorum can basically take over from there, and people delete, or the former participants that created the new quorum can delete their old shares. And that way, you can actually move devices or more specifically, bring into the quorum new devices that have new keys, so you’re not transferring key material, but generating new key material for the new devices.
So that’s pretty cool because you can essentially move your wallet around without moving the UTXOs, and that can save quite a bit of – well, moving UTXOs is expensive, right?
Mike Schmidt: It sounds somewhat magical given the current state of things. What is your assessment on the maturity of the FROST ecosystem; is this ready for primetime yet?
Mark Erhardt: I’m not sure about that. I mean, these new crypto libraries, they need to really be checked out. I don’t have an opinion on it because I haven’t looked at it myself.
Libfloresta library announced
Mike Schmidt: The next piece of software that we highlighted this week was Libfloresta. And so, this is based on Floresta, which we covered back in Newsletter #247, which is a utreexo-backed server that is Electrum protocol-compatible; we talked about that a few weeks ago. And the author of that Floresta software modularized some of the underlying libraries that he was using and published them publicly as this Libfloresta Rust library, that folks can use to achieve some similar functionality in their own wallets by using that library. So if you’re looking to do utreexo-type stuff, take a look at this library. Murch?
Mark Erhardt: All good, go on.
Wasabi Wallet 2.0.4 released
Mike Schmidt: The last piece of software that we highlighted this week was Wasabi Wallet 2.0.4 being released. And specifically, I included this piece of software as there were some new features around fee bumping with RBF and CPFP, as well as some of their coinjoin improvements and some faster wallet loading RPC enhancements and other improvements, but specifically the addition of RBF and CPFP I thought was interesting. Murch, any thoughts on the Wasabi release?
Mark Erhardt: No, sorry!
Core Lightning 23.08rc3
Mike Schmidt: Moving on to the Releases and release candidates section of the newsletter we highlighted Core Lightning 23.08rc3. We covered some of the pending features from the rc2 last week, and we also had Lisa and Dusty on from the CLN team on #262 podcast, where we chatted through a bunch of the CLN PRs that went into this release. And I’ve also seen that since this newsletter, this release is actually officially out, so I suspect we will cover it more in next week’s newsletter. So, Murch, did you have anything to add, or do you want to talk more about this next week?
Mark Erhardt: Next week sounds good.
Mike Schmidt: The other release from this week was HWI 2.3.1, which is just a minor release. It has some decoding that’s used by the BitBox02 hardware signer, and it also fixed some issues with the Qt GUI.
Bitcoin Core #27981
Moving on to the Notable code and documentation changes, I will take a moment to solicit from the audience any questions or comments you have on this newsletter. We’d be happy to take that at the end of the Twitter Space. First PR this week is Bitcoin Core #27981 and, Murch, I’m going to lean on you for this one about a bug that could have potentially caused two nodes to be unable to receive data from each other?
Mark Erhardt: Yeah, so this fixes an issue where apparently if two nodes were both trying to send a boatload of data to each other, they could get into a stall where they would not want to receive data until they had sent their own queue. And this was, I think, mostly for dust protection. Now, the behavior has been changed a little bit to prevent this stall. When nodes are in this situation, they will allow to receive some data, but then instead of blocking completely, they’ll permit some data to flow and perhaps that can break the stall. It does, however, if it takes just more than one send to move over the whole data, it will still insist on trying to send its data all the way, unless the other party declines.
This one sounded pretty weird to me, I had not heard about that before. It’s good that it’s fixed now. But it also sounds like it probably would have been a very uncommon situation.
Mike Schmidt: This is a bug that was originally discovered in the Elements open-source sidechain software that powers Liquid, and so they discovered that and surfaced this issue.
Next PR is to the BOLTs repository #919. It updates the LN spec to suggest that nodes stop accepting additional trimmed HTLCs beyond a certain value and a trimmed HTLC is a payment, affordable payment, that is not added as an output to the channel commitment transaction because it would be uneconomical chain; essentially, it would be dust value. So instead, amount equal to that trimmed small HTLC is allocated to fees instead. So, if the channel that needs to be closed while the trimmed HTLCs are pending, then that node can’t reclaim those funds. And so obviously, if that were to happen too many times, there would be a non-trivial loss, and so the node needs to limit the number of those potentially problematic trimmed HTLCs, and actually many implementations already have added such a limit. About two years ago, LDK, Eclair, and CLN added that limitation, and now it’s actually part of the spec as of this BOLTs PR. Any thoughts on that, Murch?
Mark Erhardt: Yeah, so we’ve talked about trimmed HTLCs in a few contexts lately. So, one was that there was a situation where when a node stopped responding while a trimmed HTLC was in flight, then it could cause the channel to be closed altogether, because the other node would of course see the trimmed HTLC timeout. And then it wouldn’t be worth it to claim onchain, but it would still close the channel in order to sort of end the HTLC. Then, this could cascade down the chain because now the next node would never hear about that HTLC and also close that channel. So, there was a recent change where, I think it was CLN, and now it decides to rather close downstream as failed and just eat the cost of that.
So, in light of this sort of behavior, where nodes begrudgingly accept to lose a little bit of funds, it does make a lot of sense to me to explicitly limit the exposure to these sorts of tiny payments.
Rust Bitcoin #1990
Mike Schmidt: Next PR from the newsletter is to the Rust Bitcoin repository, #1990, and it adds a “Small hash function as a smaller but slower substitute for existing hash implementations in Rust Bitcoin”, and the PR noted that on embedded processors it can lead to a 52% size reduction. Any comments on that one, Murch?
Mark Erhardt: In a previous life, I worked on an industrial camera with image recognition, and one of the biggest limitations was that we were working with C++98, because newer C++ required too much space on the device. So, I feel for these people.
Rust Bitcoin #1962
Mike Schmidt: In a similar vein, the next PR to the Rust Bitcoin repository, #1962, now checks for the x86 architecture hardware, and then runs an optimized set of instructions specifically for that architecture for any SHA256 operations, if it finds it’s on that particular architecture. So, optimization there. Any comments, Murch?
Mark Erhardt: No, not directly.
Mike Schmidt: Last PR this week is to the BIPs repository, #1485. And this is a PR to update, one, the draft drivechains BIPs, BIP300, and we noted in the newsletter, “The main change appears to be a redefinition of OP_NOP5 in certain context to OP_DRIVECHAIN”; and I guess somewhat related to this BIPs PR is, I also saw this week that there is an open work-in-progress BIP300 drivechains consensus level logic PR that has been opened to the Bitcoin Core repository. Murch, do you have nothing to say or a lot to say on this?
Mark Erhardt: I don’t think I have an opinion on this at this point. But we do have Max Hillebrand on right now, so maybe if you wanted to say something about Wasabi?
Max Hillebrand: Yes, hello and welcome, and thanks very much for the continued amazing work at Optech. 265 is a big number; hopefully it keeps going up a lot more. And with the latest Wasabi release, there are numerous important improvements, but one that I’d like to highlight is a trick on drastically improving or decreasing the wallet loading speed when BIP157 compact block filters are used. These compact block filters have a false positive rate, meaning that every public key that gets checked against the filter has a certain percentage of returning as a positive match, even when in reality this public key is not a part of that Bitcoin block and filter. And therefore, the larger number of keys you check against a filter, the larger the occurrence of false positive hits on the filter, and these blocks then have to be downloaded over the Tor network, which takes some time and increases wallet loading speeds dramatically.
So, a trick that we’ve now implemented with the latest release is that we have two separate synchronization processes, where in the first lazy synchronization, we only check for unused received addresses, so those where we are currently awaiting payments for; but we exclude addresses that have already received money in the past and addresses that are used. So, this means that initially the set of keys used to scan the block filters is much smaller, the false positive rate is much smaller, the unnecessary block download is much lower, and we can show the user an open wallet with an adequate balance very quickly. Then later in the background, we continue the synchronization of block filters across those set of keys that have been used in the past, because potentially there might be an address reuse payment, for example, that needs to be detected.
The second synchronization includes a lot more keys, therefore more blocks have to be downloaded, and it will take some time longer. And in some cases, the user will see his balance increase after that second synchronization is included, where new UTXOs have been discovered. So this speeds up wallet loading time quite substantially and might be important for other wallets that implement these block filters for wallet discovery.
Mark Erhardt: That’s a really cool approach. I was wondering, did you have this issue in practice that Wasabi wallets would download blocks more often than you expected; or how did you come to the point that you wanted to implement this in the first place? Because I did some calculations a while back, and my understanding is that if you are looking for about 1,000 scriptPubKeys, you should only need to download a block roughly every eight weeks. So do you just have so much more output scripts that you’re tracking, or…?
Max Hillebrand: Yes, that is the case specifically in a coinjoin wallet because coinjoin rounds tend to fail and we generate new addresses, and a lot of new addresses are being generated in general with these coinjoin self-spend transactions. So, yes, some Wasabi wallets have thousands, tens of thousands of addresses, and that takes quite a lot of time. And on top of that, we download blocks over Tor, which can be quite unreliable as well, and that just compounds the issues. Yeah, but it’s something, as you say, it makes a big difference if you check for 20 keys versus 10,000!
Mike Schmidt: Max, thanks for jumping on and highlighting that. I was excited about the RBF and CPFP stuff. And I saw that there was this note about faster wallet loading, but I’m glad you were able to highlight some of the tech behind there, because I think folks may be interested in your approach if they find themselves in a similar situation.
Max Hillebrand: Yes, the speed-up transaction feature was also, of course, heavily requested, especially in recent high feerate environments, which we can expect to be there in the future, so we rolled out several fee-related improvements. Some of them we could deploy on the backend side, which is a much smarter feerate estimation. Bitcoin Core is rather lazy in its adjustment of the feerate because it does not take the mempool into account. So, we now use the mempool feerate histogram feature to see the mempool, and we pick the top certain megabytes of the mempool to see what the feerates are being paid in different parts of the mempool, and we adjust our upper and lower bound of feerate for this. So, this means that the coinjoin feerates are now much more real-time adjusted and therefore a lot faster confirmation time, while still having lower feerates in a lot of different circumstances.
On the client side, the RBF and CPFP fee bump techniques have been implemented with an important preference of this complexity that should not be exposed to users, because the UTXO set is already confusing enough and RBF and CPFP are very complex and nuanced. So user-facing, it is a simple right-click and speed-up transaction button, where it shows what fee has to be paid and it just executes it. Canceling the transaction is also available. And Wasabi simply picks in the background whichever method is for the specific case preferable to then ensure that optimal block space efficiency is used.
Mike Schmidt: Excellent. Thanks, Max. Murch, anything to add here?
Mark Erhardt: I think that’s a really cool idea to make the wallet choose the appropriate tool for the job and not put that on the user. I really like that solution.
Mike Schmidt: Thanks for hopping on, Max.
Max Hillebrand: Yes, thank you again for having the podcast. And I would refer to the Wasabi blog for some further details on the different improvements, also to the coinjoin block space efficiency, there’s some interesting stuff there too. Thanks for having me.
Mike Schmidt: Thanks to everybody for joining us today and our special guests and to always, my co-host, Murch. We’ll see you all next week. Cheers.
Mark Erhardt: See you next week. Bye-bye.