Mark “Murch” Erhardt and Mike Schmidt discuss Newsletter #227.

The Bitcoin Optech Podcast and transcription content is licensed Creative Commons CC BY-SA 2.0


No significant news this week was found on the Bitcoin-Dev or Lightning-Dev mailing lists.

Selected Q&A from Bitcoin Stack Exchange

  • Did the P2SH BIP-0016 make some Bitcoin unspendable? (0:39)

  • What software was used to make P2PK transactions? (8:33)

  • Why are both txid and wtxid sent to peers? (12:59)

  • How do I create a taproot multisig address? (15:38)

  • Is it possible to skip Initial Block Download (IBD) on pruned node? (18:58)

Releases and release candidates

Notable code and documentation changes


Mike Schmidt: We’ve had a bout of heavy news for the last month on these recaps, and I think this is the counterbalance to that. So, we have no significant news this week from the mailing list, but we do have our monthly Stack Exchange segment, where we troll through questions and answers from the Bitcoin Stack Exchange and put things that are somewhat interesting or educational for the audience into the segment this month.

Did the P2SH BIP-0016 make some Bitcoin unspendable?

The first one is an interesting one because, well, I’ll jump into it, then I’ll get into some of the nuance, but the question is, “Did the P2SH BIP16 make some Bitcoin unspendable?” And the user who answered this, bca-0353f40e, noted six outputs that existed before the script format was activated, and one of those outputs was spent before the activation of P2SH. And the question is, are those remaining five UTXOs still spendable or not? And at first, I thought that the answer was that they were spendable because I was looking and reading the BIP, and there was a section saying that it was essentially activated after a certain time. I think it was timestamp and not block height, but there was some activation based on timestamp and that the rules didn’t apply to earlier –

Mark Erhardt: Earlier transactions.

Mike Schmidt: Yeah, but it turns out, at least if I’m understanding correctly now, that actually that BIP, unless I’m misunderstanding, is incorrect and the code actually implemented activation that applied to everything except for one exception block, which was the exception block that contained the P2SH transaction that was spent before activation, or before that format was activated. Have you been following along with that in my last minute updates to this particular Stack Exchange entry in the newsletter?

Mark Erhardt: I have not. I read over it once but I looked into this myself now because I was curious, and what I understand now is the rules were originally implemented in the BIP that they would only apply after activation. So, all the transactions that happened before activation would have to be validated per the old rules, and all the transactions after activation have to pass the new rules. So, any UTXOs that existed at that time would not be exempt. And so, originally when I read the BIP, I thought that they had explicitly exempted all the UTXOs that were created before activation, and would make the UTXOs’ spending be validated per the old rules, but that’s not what was done. It was explicitly applied to the transactions. So yes, they did make six UTXOs that already existed at that time unspendable.

Mike Schmidt: Okay.

Mark Erhardt: Oh, and the second part as well. The second part is, in 2018, the change was backported to the genesis block because there’s only one single exception where one transaction used this pattern before BIP16 was activated. So, they ported the rules back to count for all transactions since genesis block, and hardcoded this single transaction as an exception that is validated differently. So, that simplifies, of course, transaction verification because you don’t have to check whether a transaction happened after a specific block, you only have to check whether it’s this specific transaction and validate that single transaction differently.

Mike Schmidt: So, how do you feel about breaking those UTXOs?

Mark Erhardt: I mean, it could have been done a little more benignly by applying it to the UTXOs instead of the transactions, so that any UTXOs created before activation would fall under the old rules. But let’s be honest, 0.044 Bitcoin in what was it, 2012? In 2012, that was like 40 cents or so, so I understand why there was not extra effort done to take care of that.

Mike Schmidt: Are there other examples of making UTXOs unspendable that come to mind in a similar way; obviously you can burn your own coins, or whatever, but something that was a change to the software protocol?

Mark Erhardt: I mean, I guess I think with segwit activation, if there had been any transactions that followed the new magic pattern of having just a zero – actually, I haven’t looked into this. I think that it could have potentially happened with segwit activation, wrapped segwit specifically, but I believe that the pattern would have been nonstandard before already. So, it’s highly unlikely that somebody would fit exactly that pattern. I’d have to look into it!

Mike Schmidt: Instagibbs, I think, has something to say.

Greg Sanders: Yeah, so I was looking at the history of standardness a few weeks ago, and I’m wondering how were these outputs even made, because prior to P2SH, or what was supposed to be OP_EVAL, which was swapped out at the last second, if I remember correctly, no outputs were standard except for very small types, right? So it was like P2PK, P2PKH, and I think naked multisig up to 2-of-3, or something; everything else was non-standard. So I was just wondering, does anyone know how these were made?

Mark Erhardt: I mean, it was probably also a lot easier to mine at home, so people might have just created them, included them in their own blocks.

Greg Sanders: Yeah, I’d be interested to double-check. Maybe I’ll double-check that, but it could be people’s self-mining these patterns, I guess.

Mark Erhardt: I mean, you can still create non-standard scriptPubKeys, even without addresses. Well, that’s more of a question later in our segment. But yeah, you’d have to either convince a miner to include it. I think maybe some – was there even mining pools yet in 2012? I don’t think there – there may have been just at the time when the first few mining pools came up.

Mike Schmidt: I was looking at the block heights of these outputs too, and it seems like it’s relatively close to the activation. So, it was likely some developer who was aware that that was going to be the pattern messing around, would be my thought.

Mark Erhardt: It’s 38,000 blocks, right? No, sorry, 3,800. Yeah, that’s only three-and-a-half-weeks, you’re right.

Mike Schmidt: Well, that’s an interesting one, I thought. Anything else, Murch, you’d like to say about the Stack Exchange question, other than rejecting my edits that I attempted to make on it?

Mark Erhardt: That wasn’t me!

Mike Schmidt: I was incorrect! I thought I was fixing something because the BIP didn’t match, but I guess the BIP doesn’t match reality.

Mark Erhardt: Yeah, I think that was the author of the answer themselves.

What software was used to make P2PK transactions?

Mike Schmidt: All right, the next question from the Stack Exchange was, “What software was used to make P2PK transactions?” And this was definitely before my time, so I don’t have any firsthand experience of that, but Pieter Wuille pointed out that those outputs were created from the original Bitcoin software, early versions of the Bitcoin software, regarding the coinbase transactions. And there was also this notion of paying to IP address, where you would plug in an IP address that would respond with a public key, and you would pay to that public key after getting that information back from an IP address query asking for that public key. Murch, are you familiar with any of this ancient history?

Mark Erhardt: Only by tale and lore, honestly. So, there never was an address format for P2PK. P2PKH was sort of the introduction of addresses for P2PK, and that still worked for a long time as the coinbase output because when you write your own transaction, of course, you know the scriptPubKey already, you can just write the scriptPubKey in, so there is no need to transfer it to another user in the network. But transferring a pubkey to another user, there never were any rails for that, except this abominable idea that you ping a node at a specific IP address and ask them for where to pay, “Hey, give me a scriptPubKey”, or actually, “Just give me a pubkey”. So that’s, of course, completely unsafe because a man-in-the-middle attacker could just intercept that call, especially since none of that was encrypted, and go like, “Hey, I’m the recipient, just pay to this pubkey” and return that. That was, from what I read, removed already in 2010. So, ancient times.

Mike Schmidt: I think people are familiar with putting in an address and sending, and that being the “destination” of the bitcoins. So during these paid to public keys, what would be done if, say, you and I are communicating out of band and you provide me your public key; how would I go about sending to that? Do I craft the script myself there, or how would that have worked back then?

Mark Erhardt: I guess if you gave the pubkey out of band, you would have been able to just use send and the pubkey. That should have worked, yeah. But well, there was no checksum, there was no nothing to make sure that everything went right in the transfer.

Mike Schmidt: Satoshi Nakamoto is joining us.

Mark Erhardt: That sounds like fake news to me!

Mark Erhardt: Maybe Satoshi Nakamoto wants to comment on the original Bitcoin software. Satoshi Nakamoto, do you have a comment on some of the Stack Exchange items we’re talking about?

**Satoshi Nakamoto: **Yeah, I just wanted to say that if you really dig into Bitcoin’s early, early code, it’s script kiddie stuff. So, where it really started was, how do you make a coin; what the fuck is a coin online? And it came to me that it would be already there because the casinos were using like chips. So, that’s where the base code started in that direction.

Mike Schmidt: Thanks Satoshi. Murch, anything else on early P2PK transactions?

Mark Erhardt: Well, maybe let’s just mention that P2TR now is basically also a form of P2PK, because in P2TR, the address encodes a witness program that is de facto just a public key, tweaked with a script tree, so we’re making the full circle, we’re back to just paying for public keys.

Why are both txid and wtxid sent to peers?

Mike Schmidt: Yeah, exactly. Next question from the Stack Exchange was, “Why are both txid and wtxid identifiers sent to peers?” And this is also answered by sipa, who references BIP339, and essentially explains that passing around wtxids is better due to malleability and other reasons, but due to backwards compatibility, we still need to keep track of what the txids are in case we have a peer that doesn’t support wtxids.

Mark Erhardt: So the problem here is, of course, that the txid is only calculated from the non-witness data in the transaction, and if you want to cover the witness data too, you need to use the wtxid. So, when you announce a transaction to peers with the txid, if the node accepts an announcement via txid, a malicious peer could use that to give them a fake witness, and you’d be unable to see whether or not they gave you the actual transaction or just a malleated transaction because the witness data is not covered by the txid. So, you’d have to download the same transaction again from other people announcing it, and you might actually even be given a broken transaction multiple times if you are only asking for the txid.

So, what we wanted was to be able to request transactions per wtxid, but since that is a change in the protocol behavior, we only had that behavior in newer clients, and that’s what BIP339 is about, how to announce that you would like to communicate about transactions per wtxid and how to request them, and that only nodes that are capable of doing so will actually request them that way and trade them.

Mike Schmidt: So, there was this period of time then where segwit was activated, but the way that you were requesting transactions at the P2P layer was still txid and not wtxid, until this BIP339?

Mark Erhardt: Yeah, actually for almost three years, I think.

Mike Schmidt: Cool. Anything else on that, Murch?

Mark Erhardt: I think that covers it.

How do I create a taproot multisig address?

Mike Schmidt: Next Stack Exchange question is, “How do I create a taproot multisig address?” And so, the person asking this question saw the 998-of-999 tapscript multisig transaction and wanted to mess around and create their own multisig and do something more than the previous max key limit, that was due to some, I think, script limitations that you could do, and wanted to do this in taproot and was wondering how to do that. And so sipa again answers, and he explains that you can do this now. Previously, you were doing createmultisig and addmultisigaddress to create a non-taproot multisig, and that does not support these new wallets and only supports legacy wallets. And so, as of 24.0, you can use descriptors and newer RPCs like deriveaddresses and importdescriptors. Then there’s this new descriptor, multi_a, which maybe, Murch, you can enlighten us on. And then you can create a taproot-compatible multisig script using a combination of those RPCs in that descriptor.

Mark Erhardt: I want to correct something quickly. Descriptor wallets have been in 0.21 and the taproot descriptor exists since 22.0, but now in 24.0 the taproot descriptor is capable of using this multi_a. So, descriptor wallets are not quite as new as it just sounded. Regarding the multi_a, I’m not 100% sure, but I think that it will create a single leaf with an OP_CHECKSIGADD construction, but it might be even smarter and actually build a tree. Although for 21 out of 210, I’m pretty sure that the OP_CHECKSIGADD construction would be the cheapest.

Mike Schmidt: Yeah, you’re right. If I was inferring that the descriptor support was new, that’s not true. It’s the multi_a descriptor support that’s newer and enables the construction of these taproot-compatible multisig scripts. And yes, I don’t know behind the scenes whether multi_a is creating a giant tree or if it’s doing the CHECKSIGADD exercise.

Mark Erhardt: So basically, I’m just going to roll with that it must be an OP_CHECKSIG leaf. And what I would expect here to happen is that you basically just get this taptree with a single leave that has the 210 public keys in the script and then requires 21 signatures from those.

Is it possible to skip Initial Block Download (IBD) on pruned node?

Mike Schmidt: Last question from the Stack Exchange is about the ability to skip IBD on a pruned node. And this user who’s asking this question, Carlos, I think maybe misunderstood what the assumevalid option is. It appears, based on the sample code that they provided and the question that they were trying to do, an assumevalid for a somewhat recent block and was assuming that assumevalid was just going to not require any IBD, based on whatever block he put into that option, and was attempting to skip the IBD that way. But that is not what assumevalid is, although there’s a similarly named effort underway, called assumeutxo, in which you can download the UTXO set and validate it against a hash that would be in the Bitcoin Core source code, and you’d get a similar effect as to what Carlos was thinking that assumevalid was doing.

So sipa again, I think he answered four of these this week, points out that project. And that’s not something that’s available currently, but that’s something that’s being worked on. Murch, assumeutxo, assume valid, IBD?

Mark Erhardt: Yeah, so the IBD, or the synchronization to catch up with the current chain tip, is a point of friction when people first start using Bitcoin. And depending on what hardware you have available, that can take a few hours to two days to perform, so there’s a few cases in which you might want to skip that. So for example, if you have already performed a full synchronization yourself and you want to bring up a second node, you might trust your first node to have done the work correctly and want to transfer the state from that first node to the second node. Or, a little more dangerously, you might trust another person to give you a copy of their state and import it. So, for a while, there were ways to do that on P2P-sharing networks like BitTorrent and bootstrap a node more quickly that way.

The big danger with that is, if somebody gives you a fake UTXO set that has additional UTXOs, they could, for example, give you a transaction that spends these UTXOs or even a block that confirms the transaction that spends such UTXOs, and you would be alone on a chain tip that accepts this block, while everybody else would reject it because they don’t know the UTXO. Or even worse, they could just remove some of the UTXOs and once any of them get spent, you reject the actual blockchain and are also isolated from the network. So, this opens up nodes to a network-splitting attack, or to basically put them into a position where it’s easier to cyber- or eclipse-attack them, and then I think there’s just this big ethos in Bitcoin, where we expect every full node to have done their own full validation, because that’s where the full comes in; that’s for a fully validating node, right? So, if you just trust someone to give you a set of UTXOs, that’s sort of dangerous.

I think this is a little better when it’s just hardcoded in a release to say like the state of three months before the release, and you’re already trusting the developers not to have put in any malicious software into the release, so you might choose to also trust them that they have correctly compiled the UTXO set at a time three months before the release, and only need to catch up the last three months. But since we want to minimize trust, we actually would have implemented this in a way where even if you do this jumpstart, and you start from the point three months ago and build the remaining UTXO set from there, we would, in the background, have the node do a full synchronization from the genesis block, and then validate over time that it actually started out at the correct point. And if not, probably either warn the user or at least overwrite the old state with the actual correct state.

So, this project’s been underway for a few years already. I think that it is a pretty big lift, and this might not have been dealt the warmest reception because some people are worried that it undermines this ethos of full validation to some degree. Yeah, sorry, I rambled a bit.

Mike Schmidt: No, that’s good, that’s good background. I think the only thing that maybe you could elaborate on is, so what does assumevalid do then? You kind of got into the background and the motivation of assumeutxo and bootstrapping a second node based on the trust and source, but so what should we be using assumevalid for then?

Mark Erhardt: Oh yeah, good point. So assumevalid is an optimization for the IBD that allows you to skip script validation up to a certain point. Essentially, it makes the assumption that if a certain block header is part of your best chain, then up to this block, the scripts are correct, otherwise the blocks wouldn’t have been chained together in the best chain, because it would have been invalid. But you still parse the whole blockchain and you build your own UTXO set and you update it with each block. So, you end up having done a full validation of where all the money was created and how it was transformed over the years, but you don’t check every single script; and that, I think, speeds up synchronization by about, I’d say probably 30%, 40% or so, from the top of my head. Yeah, so it only skips script validation.

The height and the block up to which this is done is updated with each release and set to, I don’t know, like a month before the release. And so you’ll be a bit faster to validate all the blocks up to that point, but then do full script validation from that point on. You will also do full script validation if the designated header is not part of your best chain, or if you turn it off.

Mike Schmidt: Yeah, I was going to point that out as well. Actually, it’s part of the release process to update that variable in the source code. It’s part of the checklist to getting a release out. Thanks for the background there, Murch. That’s it for the Stack Exchange. Moving on to the Releases and release candidates.

LND 0.15.5-beta.rc2

LND 0.15.5-beta.rc2. Wow, that’s a mouthful. That’s just a maintenance release for LND with some minor bug fixes. I don’t think we need to get into any of that. Murch, any comments there?

Mark Erhardt: I think the only interesting one was that it uses P2TR on sent coins now for the change outputs. So, if you send money out of your LND node on chain, it will now create P2TR change outputs.

Mike Schmidt: Oh, nice, okay.

Mark Erhardt: I think the other ones were just bug fixes, as far as I could see.

Mike Schmidt: Okay, I just assumed it was bug fixes, so I didn’t even drill into it. Thanks for unearthing that for us.

Core Lightning 22.11rc2

Core Lightning 22.11rc2, a release candidate for Core Lightning (CLN), and they’ve changed their version numbering scheme to be consistent with the semantic versioning numbering scheme, which Bitcoin Core has done as well, right?

Mark Erhardt: I think they are actually doing a different thing. Where Bitcoin Core is only counting up and we increment the major version just by one every major release, I noticed here that they use something that looks like Ubuntu release numbering. So, I interpret that as year.month, 22.11, but I haven’t dug into it.

Bitcoin Core #25730

Mike Schmidt: In terms of code changes this week, there’s four of them. First one, Bitcoin Core #25730, and that changes the listunspent RPC. It adds a new argument that allows the option to include any immature coinbase outputs. So the listunspent RPC returns a list of unspent transaction outputs that the wallet knows about and lists those. And immature coinbase outputs are recently mined coins that need 100 confirmations in order to be spent, and right now, listunspent RPC does not include those outputs that are immature coinbase outputs, in contrast to something like the listtransactions RPC, that does include immature transactions. So, there’s an additional option now for the listunspent RPC to say, “Include any immature coinbase outputs”, and that’ll show up in the list.

Mark Erhardt: To clarify, listunspent only lists the wallet’s UTXO pool, so not all unspents in the network. So in this case, the immature coins would have been coins that the wallet has mined to itself. So, I don’t know exactly how furszy got the idea, but maybe some miner asked, “Hey, we want to keep track of the funds that we mined ourselves”, and for them this argument would be useful. Immature coins, of course, are the outputs of coinbase transactions which are unspendable for 100 blocks, well, only spendable in the 100th block after the coinbase was created.

Mike Schmidt: Yeah, I saw I think it was BDK that originally opened the issue for this, so I’m not sure if that has anything to do with Stratum v2 or mining or anything like that. But yeah.

LND #7082

LND 7082. So, you can create no amount of private invoices, but in LND, if you didn’t provide an amount, then routing hop hints were not provided. And if you did provide amount with the private option, it was providing the route hints. So, the PR here allows route hints to now be provided for private invoices that don’t have an amount provided. And as a quick background, route hints are data that suggest part of a routing path between a spender and a receiver, so that they can send payments through nodes that potentially they previously didn’t know about. And so for whatever reason, those route hints were not populating for a private no-amount invoice creation in LND previously.

Mark Erhardt: Maybe to clarify, this should not be confused with blinded paths. So there is a feature in Lightning that allows the recipient, instead of telling the sender about their own node, to essentially give a package of a few layers of an onion that declares an endpoint that if the sender reaches, will have a route included to reach the recipient. And the route hints, on the other hand, are basically just a set of directions how to find a hidden node, or I should say an unpublished node, in the network. And yeah, for no-amount invoices, there’s also the problem, or the added difficulty, if the last hop next to you notices that the invoice might have not had a specific amount declared, they can keep more and forward less and steal money that way. So, yeah. Anyway, now we have some hop hints for no-amount invoices in LND.

LDK #1413

Mike Schmidt: Next PR here is LDK #1413, and that removes support for the fixed-length onion data format. We’ve covered this in a few different newsletters. BOLT4 was updated to remove the fixed-length onions from the spec and that was about a month ago. And actually, even before that was removed from the spec, CLN, LND, and Eclair also had previously removed support for that. So, it seems like this is the last domino of that to fall. Murch, thoughts?

Mark Erhardt: Yeah, I think we’ve talked about this quite a bit already, so we might not want to bring it up in the next one anymore! It’s just this old format that was less flexible and basically over-defined is now removed completely from the network. We’ve already reported that something like all but three nodes on the network weren’t using it anymore, so I guess those last few nodes should maybe, at some point, update from their well-beyond end-of-life versions of software, and I think we’re smooth sailing from here.

HWI #637

Mike Schmidt: Last PR this week is the Hardware Wallet Interface (HWI) #637, and it adds support for the forthcoming Ledger app 2.1.0, and that’s the software that supports the Ledger signing devices, hardware wallets. And in that Ledger app, there’s additional support that comes for allowing multiple internal xpubs, support for the wsh descriptor, Witness Script Hash descriptor, support for signing messages, OP_RETURN output support and wallet policy language. So, there’s a bunch of goodies in the Ledger software that HWI now can support once Ledger app 2.1.0 is released, which I don’t think it is yet. And the only other thing here is, the author who made these changes to HWI, Salvatore, who spoke with us last week, is also working on a BIP for manuscript wallet policies, which is cool.

Mark Erhardt: I think at this point it’s also fair to give another shout out to Andy. Andy has, over the last few years, built this library for Bitcoin Core, the HWI, and I think it has been one of the driving contributors to the wallet interfaces, especially hardware wallets, to get a lot more compatible. He’s the author of PSBT, the HWI, and the Output Descriptor Standard. So, I think these are three of the key points that make it much, much easier for wallets to interact and create multiparty transactions and multisig transactions with each other. So, I think I just wanted to mention that this is a really important effort and if you see Andy, you should be appreciative of that.

Mike Schmidt: Yeah, a shoutout to achow for all his great work. Good to have you back, Murch, thanks for joining. Thanks everybody for listening and we’ll see you next week.