Episode 68: RBF In Bitcoin Core 24.0
Ep. 68

Episode 68: RBF In Bitcoin Core 24.0

•

Episode description

Aaron and Sjors revisit replace-by-fee (RBF). As they mentioned in episode 65, the upcoming Bitcoin Core release — Bitcoin Core 24.0 — includes the option to switch on “full RBF”, but this has caused some commotion in the Bitcoin community since the recording of that episode. Aaron and Sjors explain what this commotion has been about, and they highlight some of the new arguments for and against (full) RBF.

RBF has been the topic of a previous Bitcoin, Explained episode: episode 26. In this new episode, therefore, Aaron and Sjors don’t explain in-depth on what RBF is, exactly, or how it works. They do however very briefly summarize its most important aspects.

Aaron and Sjors then go on to explain why Bitcoin Core developers originally decided to include this feature, and they discuss some of the arguments for and against (full) RBF that came up at the time and since then. These include the effect of RBF on “pinning attacks” (a type of attack that is especially relevant for the Lightning Network and other Layer Two protocols), the relative safety of accepting unconfirmed transactions today, privacy-related arguments concerning the “opt-in” flag that RBF transactions currently use, the detrimental effects of monitoring the network for potential double spends, and more. Aaron and Sjors also discuss the pros and cons of including RBF as an optional feature and thus letting node operators decide for themselves how their node deals with conflicting unconfirmed transactions. Sjors outlines why, in some cases, giving users more options could have detrimental effects on the health of the Bitcoin network, and considers whether the option to include the RBF option is such a case. Finally, Aaron and Sjors briefly discuss an initiative by full RBF advocate Peter Todd to incentivize miners to apply full RBF logic to their transaction selection.

Chapters

00:00:00 Full RBF and the Bitcoin Core 24.0 delay
00:06:01 Pinning attacks and the case for full RBF
00:11:25 Lightning complexity and miner incentives
00:15:59 Muun Wallet and zero-confirmation payments
00:18:53 Merchants and the risks of accepting zero-conf
00:21:46 Privacy, stuck transactions and merchant trade-offs
00:26:19 Removing a merged feature: release process and debate
00:28:39 Node options, diverging mempools and mining risks
00:36:36 Peter Todd’s full RBF miner bounty

Transcript

Download transcript (.vtt)
0:20

Aaron van Wirdum Live from Utrecht, this is Bitcoin.

0:23

Sjors Provoost Explained.

0:25

Aaron van Wirdum Sjors, why is Bitcoin Core 24 not released yet?

0:28

Sjors Provoost It's almost released.

0:30

Aaron van Wirdum What are we waiting for?

0:31

Sjors Provoost But there was a bit of a snafu. There was a bit of discussion about SBX, I mean RBF.

0:36

Aaron van Wirdum Aha! The discussion about Replace by fee.

0:43

Sjors Provoost That's right.

0:44

Aaron van Wirdum Oh, it's such a coincidence you bring that up, because that's what we're going to discuss in this episode, Sjors.

0:48

Sjors Provoost Exactly, I thought so.

0:51

Aaron van Wirdum So yeah, there's been a long discussion about RBF. It's not the first time, but it definitely resurfaced now that Bitcoin Core 24 is coming up because there is a new feature included in Bitcoin Core 24, as we explained in our episode about Bitcoin Core 24, which was two episodes ago.

1:12

Sjors Provoost Yep.

1:12

Aaron van Wirdum Episode 60, no, three ago, I think.

1:15

Sjors Provoost Could be several episodes ago.

1:17

Aaron van Wirdum A couple episodes ago.

1:18

Sjors Provoost So we talked about how fast release candidates usually go through. And this time it took longer. And the reason it took longer is because this feature was in there.

1:27

Aaron van Wirdum Right. Well, this is the first time I think that a Bitcoin Core release candidate has been delayed because of a feature, I think, right? Because usually it's to find bugs. But this time this was like explicitly meant to be in there. But then discussion happened whether that was actually desirable.

1:44

Sjors Provoost I don't know if it's the first time. It's definitely the first time I remember that there was a lot of discussion about whether something should or should not be in the next release.

1:52

Aaron van Wirdum After it was included in a release candidate.

1:54

Sjors Provoost Yeah, typically you want the release process to be pretty boring. So there is a release candidate, people look for bugs, and then there's another release candidate and people look for more bugs. But this can happen if there's a feature in there that's controversial. Then you can have discussion over it just means that it's not going to be a release for a while.

2:12

Aaron van Wirdum And by the way, I'll claim my… I was right this time. I'm not sure did I mention it in the episode? I definitely mentioned it before the episode. I told you this is going to be an issue, and then you said no it's fine and I said no this is going to be an issue and now it became a huge issue. Sure. I was right about this one.

2:29

Sjors Provoost You may have been, I don't remember.

2:31

Aaron van Wirdum I definitely I remember very well. Anyway, so it's Sjors, what's the feature?

2:37

Sjors Provoost The feature it's called mempoolfullrbf it's a setting. And this setting what it does is instead of when you have transactions that signal replaced by fee you will accept those transactions, if they pay a higher fee and if they obey some other rules that were specified in BIP 125, and if you turn on this particular setting then any transaction that does not signal this flag, but does comply with all the other rules from BIP 125 will get replaced. So the word full is a bit misleading. It's definitely more RBF, but it's, there's still rules. You can't just replace any transaction with any other transaction. There's still some rules to apply there.

3:21

Aaron van Wirdum Yeah, we made a whole episode on replace by fee, RBF. So we're not going to repeat what it is exactly. But very briefly, you can replace a transaction if you include a higher fee, right? That's essentially what it means.

3:35

Sjors Provoost It has to pay a higher absolute fee. It has to pay a higher fee rate and a few other conditions. Yeah, we've explained it in episode 26.

3:43

Aaron van Wirdum Right. So usually it's already possible. Now you include a flag in a transaction where you basically tell the network, look, I might want to replace this later. And to be clear, technically a miner can always replace a transaction, right? But it's just with this flag, you make it very explicit and then nodes will also forward it to other nodes.

4:04

Sjors Provoost But that's the main thing this setting does. It's about what it forwards and what it doesn't forward. So if you have a direct connection to a miner, then... That miner can do whatever it wants because this is not a consensus rule. Right and in fact there's already a bunch of nodes out there on the network that run maybe a version of Bitcoin Knots or some other patch that already have this setting in there so it's not really a new feature but it is new that that it's available in Bitcoin Core. It's off by default right?

4:31

Aaron van Wirdum Yeah. So now it's a new option in Bitcoin Core and it's off by default how do you switch it on?

4:38

Sjors Provoost You put a setting in your config file called mempoolfullrbf

4:42

Aaron van Wirdum Okay that's it so what are we gonna do in this episode, Sjors I think I know we're gonna discuss what the more recent discussion has been about more or less.

4:52

Sjors Provoost Yeah I think we're gonna try and not rehash the entire RBF discussion that's been had historically but maybe try to focus on the things that were new in that discussion but we'll probably get to some old stuff too.

5:04

Aaron van Wirdum Yeah. And also, so this is one of these debates where there are kind of several arguments on both sides. And then all of these arguments have... Counter arguments and then these counter arguments have counter counter arguments and then these counter counter arguments have like sub arguments.

5:26

Sjors Provoost And also there's just six six or seven whatever, lines of argument that go in parallel right? So, you can talk about very different aspects of this feature and also of this discussion which we'll get to.

5:38

Aaron van Wirdum Yeah there's like all of the sub arguments and counter counter arguments like there's also… they… can all be summarized by like a meta argument. And there's there's a lot of sides to this discussion, and I don't we're definitely gonna forget or miss some of them but hopefully we'll get the most important ones at least during this episode. Either case we'll do our best. Sure, where do we start? Why was it included? Let's start there. Why was this feature included in this version of Bitcoin Core? What was the original argument?

6:10

Sjors Provoost Because there was a pull request. But yeah, basically there was a pull request that argued that there are certain annoying pinning attacks, which I guess we'll try to explain, in Lightning and other protocols that are multi-party. So these could be vaults or even CoinJoins perhaps.

6:29

Aaron van Wirdum Yeah, sort of smart contract-y kind of stuff.

6:31

Sjors Provoost And basically these protocols are being, the argument was these protocols are being burdened by some of the details of the RBF rules. And one way to reduce those burdens at least is to have this new full RBF option. That was the argument.

6:50

Aaron van Wirdum Okay, hang on. What's a pinning attack?

6:52

Sjors Provoost Well, this is where things get really complicated. And this complexity is something we'll get back to because we are not the only ones who don't fully understand these things. But basically... What happens is if you want to replace a transaction, you have to obey certain rules. And just the simplest rule of that is I think the new replacements can only be 25 transactions. Maybe it's 100. I don't know the number. But that could be the rule. So now if you have a situation...

7:22

Aaron van Wirdum I think it's 100, by the way.

7:23

Sjors Provoost Could be 100. So now let's say there's a situation where I, the attacker, have created 101 transactions, that are designed to annoy my victim somehow and the victim wants to replace those by paying a higher fee but unfortunately the victim cannot replace 101 transactions because you can't replace more than so the victim cannot replace 101 transactions because they can only replace 100 transactions so that class of attack where the victim cannot replace even though they might have the fees is called a pinning attack and other things that you can do is

7:56

Aaron van Wirdum Let's say this is a very specific example of one particular pinning attack then I I would assume right?

8:02

Sjors Provoost Yeah the there's all sorts of ways that you're essentially you're pinning a transaction that you want as the attacker you're pinning that into the mempool making it hard to unpin it I guess.

8:11

Aaron van Wirdum Right so then in that case, yeah as the attacker you wouldn't add the RBF flag obviously because you're trying to do it to annoy someone, but then if full RBF would be…

8:23

Sjors Provoost Or you do. I mean, for the example I just gave, with 101 transactions, you can mark them RBF, but when somebody then tries to replace it, they still cannot replace it because one of the rules is no more than 100 transactions.

8:37

Aaron van Wirdum Right, but in that case, the full RBF rule would not fix this.

8:41

Sjors Provoost No. But there's other attacks where you can turn RBF off as an attacker. So the official specification of RBF, of opt-in RBF, says that... If your parent transaction is RBF, then your child transaction should also be RBF. But that's not being checked, that rule. So now it's possible to create a transaction and then the attacker would put a non-RBF transaction on top of the RBF transaction and then you can't replace that anymore because the second transaction does not signal RBF. And that one might be fixable with this change, right? Because then it doesn't matter whether or not the transaction is RBF as long as it obeys the other rules.

9:22

Aaron van Wirdum Right. So in that case, it actually sounds like the problem with full RBF... Wait, let me rephrase that.

9:30

Sjors Provoost Opt-in RBF?

9:32

Aaron van Wirdum I was going to say, yeah, the problem... Whatever problem we're talking about here can be solved with... The problem with full RBF in this case is that it's not RBF enough. Like there's still limitations to what you can replace.

9:50

Sjors Provoost Yeah, so in the second example I gave, you could actually, full RBF would be a solution for that problem. Because, right, so the attack would be you put a transaction that doesn't signal RBF on top of a transaction that does signal RBF. Now this new transaction is unreplaceable. But if you have full RBF then it doesn't matter that the that the second transaction doesn't signal RBF it's going to be treated the same anyway so now you can replace that one so there's some things that you can you can fix with full RBF but the example of 101 transactions you cannot fix with full RBF because RBF has other rules.

10:27

Aaron van Wirdum Yeah okay well let's not dwell on this enough let's just establish that some types of smart contract-y kind of things on Bitcoin would benefit from full RBF or at least they would benefit from maximum RBF which is the term used for even fuller than full RBF where yeah really any transaction that has higher fees would be replaced without any other rules.

10:53

Sjors Provoost This this term may have been coined by Suhas Daftuar I think his last name is… Maximal RBF would be like if miners only cared about their financial incentives that would be you know they would replace things in a certain way but it's actually not that trivial to to really think that. And there's trade-offs there when it comes to...

11:11

Aaron van Wirdum Sure. Well, where I was getting at was, so let's sort of move past this complex smart contract stuff and how it can be attacked exactly. And we'll assume that RBF helps because that is at least what Lightning developers think.

11:26

Sjors Provoost Well, this is where part of the confusion started coming from. So basically from a Bitcoin Core point of view, there is a pull request with somebody saying... Well, let's enable full RBF because it makes these pinning attacks less bad, and then it links to a bunch of mailing list posts that contain very long stories about very complicated pinning attacks in Lightning that nobody outside of Lightning understands. So now, as a Bitcoin Core reviewer, you're looking at that pull request, and you're like, hmm, do I really want to read 15,000 pages of Lightning mailing list posts to decide whether this feature is good or not? You might decide not to read those mailing list posts and just assume that they know what they're talking about. And then you focus on other aspects of the discussion, but not that specific one.

12:12

Aaron van Wirdum Well, it is interesting you mention that because that is something that came up. In our last episode right with some where I mentioned or we mentioned that sort of Bitcoin development and Lightning development is kind of diverging a bit it seems yeah people that are interested in one or the other rather than both that sort of growing apart a little bit it seems sometimes so this would be an example of that where okay the Lightning devs say that this would actually benefit their protocol and then what you're saying is that a Bitcoin Core developer might think: All right, I believe you.

12:45

Sjors Provoost Right. Well, in this case, those Lightning developers probably do understand the Bitcoin Core side of things on this aspect. But most of the Bitcoin Core developers will not understand the Lightning side of things.

12:54

Aaron van Wirdum Right.

12:55

Sjors Provoost So it's a bit asymmetrical. So there's a little bit of trust there. So then there's still other reasons why you may want to or may not want to merge these things. One of the things that came back in a discussion now recently was… well these pinning attacks are not really solved. So then should that pull request have been merged in the first place given that it didn't really solve anything and it did piss off other people. So that's I guess we can get to that.

13:23

Aaron van Wirdum Yeah well okay so so I was gonna go one step at a time. So first of all at least it seems that full RBF would solve a problem for Lightning whether or not that is really the case we can sort of park for now but that was at least the initial assumption. Yeah and then I think the second big assumption or that's not even an assumption maybe but is that full RBF is incentive compatible at least if you look at if you assume that miners want to maximize profits they'll in the long term anyways, include whatever fees earn them the most money, right?

14:09

Sjors Provoost That's the argument. Yeah, it's more incentive compatible because it doesn't require miners to not mine something that would make them more money in the short run.

14:18

Aaron van Wirdum Yeah, so as a Bitcoin Core developer, you're hearing from Lightning developers that this would help them. And then you're looking at the incentives and you figure, well, it's also incentive compatible, at least on the long run. So let's include it. That's kind of...

14:32

Sjors Provoost I think that was most of the discussion. Yeah.

14:34

Aaron van Wirdum Most of the argument for including it anyways.

14:37

Sjors Provoost Yeah.

14:37

Aaron van Wirdum Right. Okay. Does that include, so have we covered the reasons to include it? We didn't miss anything here, right?

14:44

Sjors Provoost We may have missed something, but I think those were the two, the two main things. Yeah.

14:47

Aaron van Wirdum Yeah. I guess you could maybe argue that a third argument. So the third argument is maybe more about timing. So this became part of the discussion. Like even if everyone agrees that long-term RBF is incentive compatible and therefore probably unavoidable, you can still ask the question, should it be included now? Should it be phased in? Should it be announced that it's going to be the default in a year from now? How do you make the transition from not having it in Bitcoin Core at all to having it in Bitcoin Core?

15:23

Sjors Provoost I think that discussion did not happen in the original pull request. And that's kind of what flared up afterwards once people found out that this pull request existed. Because, you know, just because a couple of people in Bitcoin Core decide to merge something doesn't mean that the entire Bitcoin ecosystem knows that this happened. It can take a while. It took several months, in fact, for some people to find out who, basically people who rely heavily on zero confirmation transactions and who knew this time was coming at some point, but, you know, they may not expected it to be now.

15:59

Aaron van Wirdum All right so why did why did the discussion originally flare up like what was the original argument against what we just laid out?

16:08

Sjors Provoost So there are companies that rely on zero confirmation transactions. Yeah.

16:13

Aaron van Wirdum Well I was specifically so I think it flared up, it started with Muun the Lightning wallet or apparently not exactly a Lightning wallet… they first made the case that Maybe not right now anyways. And do you know the argument there?

16:32

Sjors Provoost Yeah, so my understanding of the argument, because I was not on the mailing list when that flared up.

16:36

Aaron van Wirdum How were you?

16:37

Sjors Provoost Well, I was on the mailing list, but the mailing list sits in a mailing list folder on my computer, which I check once every three months or so. So I may be behind on some of the drama. But basically, my understanding of the argument is that Muun Wallet, or Muun Wallet, I guess, the way they handle Lightning is that the user sends an… sends a on-chain transaction to the Muun server and then they send a Lightning transaction to the final destination and That itself might be fine except that they're using an unconfirmed transaction for this in order to make it fast. So the risk there for the company is that users will make a unconfirmed transaction to the Muun server the Muun server then sends the actual Lightning payment that is irreversible to the final destination and then the user simply undoes the transaction and double spends the server.

17:24

Aaron van Wirdum Well you say simply but that that would be very simple right? You you're using the Muun software so I'm assuming, well the Muun wallet won't include the feature that will allow that.

17:33

Sjors Provoost So so you would either just take the keys out of the Muun wallet put it in some other wallet and then do it that way right? Or you figure out how the Muun wallet communicates with the Muun server and you just write your own software that does that and creates a bunch of fake accounts and does it with a lot of money.

17:50

Aaron van Wirdum Yeah, I'm just saying that's not simply for people like me, maybe for people like you, but it's possible.

17:55

Sjors Provoost I would say it's for people who are sufficiently motivated to do this.

17:59

Aaron van Wirdum Yes, I agree.

18:00

Sjors Provoost Now, the way they prevent that scenario from happening is by not signaling RBF. So their transactions do not opt into RBF, which means that by default, if an attacker tries to do this and they're only connected to regular Bitcoin Core nodes, those transactions won't propagate. But that is still quite risky because, again, a sufficiently motivated attacker would find one of Peter Todd's nodes and just connect to those or directly connect to a miner if they can find it and start stealing it. So essentially it's not very safe, but at the same time it's been working and I think they know that it's not safe and they want to improve it. They just were hoping that they'd have another year on the roadmap before they would have to do this.

18:44

Aaron van Wirdum Yeah. For context, Peter Todd's nodes, by that you mean, he runs a bunch of Bitcoin Core nodes that have a patch for full RBF, right?

18:51

Sjors Provoost Yeah, and there's others that do that.

18:53

Aaron van Wirdum Yeah. Okay, so at that point, the discussion really flared up. And that's when some merchants, basically, or payment providers, payment processing type of people started to weigh in. And I think the argument can be summarized pretty easily there, which is that... I mean, this is an argument we haven't even mentioned in favor of full RBF, which is zero confirmation transactions aren't secure anyways. They're just not secure. People shouldn't rely on them. So full RBF, no RBF, flag RBF, it doesn't matter. You shouldn't trust zero conf.

19:31

Sjors Provoost Right, but there's a difference between saying that and then making the situation even worse, right?

19:37

Aaron van Wirdum Well, that's where I'm getting at. So then this is one of these things where a counter argument is a counter argument as a counter [unclear].

19:44

Sjors Provoost Yes.

19:44

Aaron van Wirdum So yeah. So then the counter argument there, the first counter argument is, well, that might technically be true, but you know, we run a payment processor and we've processed tens of thousands of payments and the number of double spends that we've actually seen is negligible. Like right now there isn't actually a problem with double spends that's you know worth taking very seriously or at least it's very manageable. And then the counter counter argument as you mentioned is… No, that's not the counter counter. See, now I'm confusing my own arguments.

20:20

Sjors Provoost Well, you can go all sorts of directions with that counter counter argument. You could say, well, but if you create an ecosystem that is too dependent on things that are unsafe, you get a strong lobby to keep those unsafe things in. And then in the long run, it goes really bad, right? Because you're building a really weak foundation. So you could argue, no, we need to actively push more discipline into the system. That's almost similar to the argument we had about... The Lightning channel closures. So the Lightning node crashes right now.

20:49

Aaron van Wirdum Anyways, the point being, you just mentioned something about sabotaging the network, right? Which argument was that? Was that some sub-argument or counter-counter-argument?

20:58

Sjors Provoost Well, the argument is you can tell a company like, hey, your business model relies on something that is inherently insecure, namely zero-conf. You can tell a company that and then it's their decision. Or you could say, no, no, no, it's not their decision. We're going to actively make them stop doing that. And that could be an argument for simply deploying full RBF, even though you have no other reason to do it. So if a pull request had been opened, and that comes back to the Bitcoin Core process, I guess, if a pull request had been opened saying, let's just enable full RBF for no other reason than to discipline these zero-conf companies, then that argument probably would not have survived the pull request process but that's not the only argument right? There's other arguments, namely it might help with these pinning attacks, yadda yadda, yadda.

21:46

Aaron van Wirdum All right, so Sjors, as I mentioned I don't want to dwell on the discussion about RBF too much but just to sort of work our way through a couple of arguments that I'm not sure I've heard before real quick so one of the arguments is that requiring an RBF flag in order to have your transactions be replaced is bad for privacy. Like it sort of reveals something about the user and it can be used to, you know, analyze the network and maybe cluster transactions, that kind of stuff. So if we can get rid of the flag that would improve privacy, that's one argument.

22:19

Sjors Provoost It's another fingerprint that tells you something about the wallet, et cetera.

22:22

Aaron van Wirdum Yeah, exactly. There are also good reasons to maybe want to revert transactions. You know, an obvious example would be you sent a transaction accidentally. You know, that's... Well, in this case, you might want to.

22:37

Sjors Provoost You sent a transaction accidentally that did not signal RBF.

22:41

Aaron van Wirdum Yes, exactly.

22:42

Sjors Provoost So that would only be a problem with wallets that don't do that by default. But yes, that's an argument.

22:46

Aaron van Wirdum Or another argument, this is kind of the standard argument, I guess, is we've seen in the past couple of days where suddenly the mempool fills up with tons of transactions. And there's definitely people on the network that don't use the RBF flag just because their wallet doesn't do it by default or they haven't upgraded their wallet for a long time and now these people have stuck transactions so it might make sense for them if they can just replace it.

23:11

Sjors Provoost Yeah maybe though then the counter argument would be it would be strange to have a wallet that does not support RBF but that can somehow replace transactions that don't support RBF. But anyway, they might switch to a different wallet just to replace the transaction, something like that.

23:28

Aaron van Wirdum Yeah, well, that's a good example. Peter Todd's argument is that right now a lot of merchants that want to prevent double spending do this by monitoring the network very extensively, which requires... Resource costs from all the nodes they're connecting to so in that sense it's bad for the health of the network.

23:47

Sjors Provoost Yeah so I guess the idea here is that the the fee bump transactions or the competing transactions could be in some corner of the network where they're trying to double spend you and you're in a different corner of the network so you don't see that you double spend so as a merchant in that case you really want to be completely up to date on the mempool more so than other nodes.

24:04

Aaron van Wirdum Yeah there's something Sergey from Bitrefill brought up which is called the the issue with the free American call option am I saying that right? I think so. Which means a merchant that receives a transaction if it's easier to replace the transaction then it could have a problem if the price becomes volatile because users might decide to change their minds about the purchase and sort of halfway just opt out so even if you do I guess… Yeah, what's our reasoning here? Even if you do wait for confirmation, then until the confirmation happens.

24:43

Sjors Provoost Well, I guess as a user, what you can do is you can send a very low fee transaction to the merchant and then the merchant is just going to wait. Well, if the merchant just waits for confirmation, then there's not really a problem. But if the merchant wants to give you the product early or something like that, then you might, well, you can already double spend. But I guess...

25:05

Aaron van Wirdum That's why I mentioned the confirmation.

25:07

Sjors Provoost I guess the other thing that might happen is that the merchant doesn't give you the product yet, but they might lock in some sort of price. Yes, exactly. So they might say, well, we're not going to ship anything yet to you, like a gift card, but we're going to wait until there's a confirmation. But if there's a confirmation, this is the price you're going to get, which is still sort of a zero-conf thing to do, but slightly less so. And then in that case, yeah, the user could see, well, if this deal... You know, because the price is already locked in, if the price moves in the wrong direction, the user can double spend it. And that's already possible if the user has opt-in RBF turned on. But then the way merchants try to prevent that is to tell users don't turn opt-in RBF on. And that, of course, won't work anymore if the flag has no meaning.

25:55

Aaron van Wirdum Yeah. And the counter-counter submeta argument there is that the merchant can also use child pays for parents. Anyway there was also something with CoinJoin but which also seems like it wasn't a very strong argument because even some of the CoinJoin

26:09

Sjors Provoost I think the CoinJoin thing was just about pinning attacks. So there was a worry about pinning attacks that could be slightly reduced by having this full RBF.

26:19

Aaron van Wirdum Right. Okay. Well so anyway, so we're rushing through a couple of the arguments and some of them which I hadn't heard before. But I think the real issue is. Should this have been included in Bitcoin Core? And now that it was included in Bitcoin Core, should it be removed before release? That's what a lot of the discussion ultimately sort of wound down to, right?

26:40

Sjors Provoost Yeah, and it was tricky because basically the discussion itself was holding the release back, but you also get a little bit into these power dynamics of once something is already merged, it takes more energy to get it out in a certain sense. Somebody has to open a pull request to get it out, and somebody has to then spend the energy defending that pull request. And so what ended up happening is somebody, I think multiple people tried making pull requests to get it out, and especially the last attempt had a lot of discussion. And we'll get to some other arguments that we use in that discussion.

27:11

Aaron van Wirdum More arguments still?

27:13

Sjors Provoost Yeah, yeah, yeah, for sure.

27:14

Aaron van Wirdum Oh, my.

27:14

Sjors Provoost But basically what happens is if you open a pull request and a lot of people don't like it, other people do like it, you get lots of toxic discussion on the pull request, at some point you might decide, like, I don't want to bother with this. And then essentially somebody can force a change or prevent the change by just creating a lot of drama, not even having any arguments. Just the drama itself can be a factor that decides whether a certain change happens or not. And that, of course, is not a healthy development because that could be used to, I don't know, prevent the soft fork from happening or... Make a soft fork happen that we don't want to happen. Some other controversial change could just be happening because you're just making you know anybody who so for example maybe you sneak in some some evil change in the code somewhere when nobody's paying attention and then anybody who tries to remove that evil change you just completely make their life miserable by by making really a lot of drama around it you know that could work maybe not there's just of course other things that could happen in that case right But here, I think it's fair to say that, or at least I think it's fair to say that the reason that full RBF is still in there is not just because the arguments weighed in one way or the other. It had a lot to do with the social dynamics of just making a lot of noise on any attempt to remove it.

28:39

Aaron van Wirdum Okay, Sjors, you mentioned that was one more argument. Actually, I think I know where you want to go with this. Okay, let me first bring up one point myself. So this is sort of my general philosophy when it comes to Bitcoin Core development or the Bitcoin Core release process or what's included or what's not included. I tend to like options. I think it's fine to give people options. I think it's fine to give users options. I think people should be... Free to run their own nodes however they want basically. Now obviously Bitcoin Core developers don't have some sort of obligation to include features, but in this case it seems pretty obvious that if some Bitcoin Core developers think it's a good idea some think it's maybe not a great idea, then I would say include it as an option that people choose. The people that will actually make that choice, you know they're the people that are running Bitcoin Core. And they know how to change a default. These are pretty technical people. And I think they should be trusted to be able to do that. Now, what the default is, that's maybe a more interesting discussion, especially in a case like this. I don't know where I would land on that discussion.

29:53

Sjors Provoost But that's not the actual discussion here, so that's fine.

29:56

Aaron van Wirdum Including it as a default, in my mind, that should always be fine.

29:59

Sjors Provoost Including it as an option.

30:00

Aaron van Wirdum Sorry, yeah. Including it as an option should always be fine.

30:02

Sjors Provoost So I agree with that in general, but I can give you a very different topic area where it's very clearly not so obvious. So, for example, there is an option called, I think, assumevalid which is the block where we, you know, which has been baked into the source code where we say you cannot, you don't have to verify signatures before that. Actually, that may be the wrong example, but they're similar. There's some options where basically if you provide a malicious, if somebody gives you a malicious suggestion for that option, they can steal your coins. So I don't think it's true for assumevalid, but it might be for assumeutxo for some of the proposals we say. So you could give the user like, oh, just add this to your Bitcoin Core settings and things will be great for you.

30:49

Aaron van Wirdum Well, if you say it in that voice, it's very compelling.

30:51

Sjors Provoost Exactly. So there are some things that you might be able to, that could be abused by making it a little bit too easy for users to configure things that would shoot them in the foot. So those settings generally are not in there. And there's definitely settings that are hidden in a way that you can only use them on test networks you can't use them on the real network. So as a general rule basically you should be able to configure whatever you want, but if it's a true footgun, then maybe not or at least it should be a little bit more difficult so that you really know what you're doing. The question is does that apply here? Now initially you might think no. It doesn't really apply here why would it matter that somebody turns on this full RBF and or not? And so this is where the last I guess… attempt to undo this pull request came from the realization by at least some people that maybe this is one of those cases where it's not a good idea to give everybody an option because it could actually have some cause some problems down the line that are completely separate from whether or not you think RBF is good or not. And this is, I believe, posted initially by AJ, Andrew Towns, and by Suhas as well as a follow up on the Bitcoin developer mailing list. And the general topic was like, do we want mempool policies to start diverging? And what this means is that if you run the mempool with one setting and I run it with another setting, then our mempools are going to be different. That are some settings where this different isn't a big deal. So for example, if I set my mempool to one gigabyte and you leave it at 300 megabytes, then my mempool is going to be bigger than your mempool. But basically the top of the mempool, so the transactions that are going to be in the next block, that's going to look the same between us. But if one of us turns on RBF and the other person doesn't, it's going to be different. Now, what this means is that if you're a miner, you want to mine the economically most efficient block. Now you're going to have to connect really to both of us because you want to see what the best block is in mining pool type 1 and what the best block is going to be for mining pool type B.

32:55

Aaron van Wirdum Don't you only have to connect to the full RBF node?

33:00

Sjors Provoost No, because they might have mutually exclusive transactions in them. Because I could... I mean, you'd have to come up with some esoteric... Examples I guess because in this case

33:12

Aaron van Wirdum Yeah I think the example there would be that a transaction with a low fee has very high fee child transactions so it's it's bumps from the full RBF node and therefore the high fee child transactions are also bumped from the high fee node like something that this would be the scenario something like this right

33:33

Sjors Provoost So I'm I'm not actually entirely sure about this one so in principle at least You know, there could be things in the full RBF node that are not in the opt-in RBF node, but it may not be the other way around. That may not be as big a deal. But still, it creates some sort of split. So the worry there is that there might be some settings where these kind of splits could happen, and they could be worse. They might not be just in one direction. There might be things that people can toggle that work in both directions.

34:00

Aaron van Wirdum Okay, so what is the exact problem here? That a miner has to connect to multiple nodes?

34:05

Sjors Provoost Yes, exactly. And that comes back to similar to the argument of having nodes connect to lots of nodes. The miners would be connecting to lots of nodes. If that goes too far, then miners would have to have a lot of connections to a lot of different nodes, which means it's more expensive to run a miner, basically. And conversely... As a miner, you probably want to be very reachable to different kind of nodes. So you might want to almost identify yourself as a miner so that all these different sub-mempools basically that are out there all connect to you. And that's not good either because you want to have some privacy as a miner. So the very long-term consequence is that if you make it too easy for users to start choosing things somewhat arbitrarily that create different kinds of mempools, then you might end up in an ecosystem where mining becomes... More difficult, therefore easier to centralize. That's sort of the long-term risk. Now, whether this one little setting really matters in the scheme of things, that's, I think, very debatable. And so that's why maybe, you know, but that was the reason why this last pull request was opened to say, well, we're worried about this scenario and we should think about this scenario a bit more. And so maybe we should do this option in the next release while we have some time to think about whether this really matters or not.

35:19

Aaron van Wirdum Yeah, I guess the counter-argument there would also be that this is already a potential long-term risk because people can always patch their Bitcoin Core node or run something else.

35:31

Sjors Provoost Yeah, but they're much less likely to do it at a scale that it matters for your revenue.

35:36

Aaron van Wirdum Maybe, maybe not.

35:39

Sjors Provoost But anyway, the hope was to have... To not have two different versions, to have a new version of RBF, maximal RBF, or transaction version number three, or some sort of proposal that everybody would be happy with, at which point you would not have two different policies. You just have one, and you'd recommend it to everyone. So the idea would be to wait a little bit and not have the split. But you could also say these type of split scenarios are not likely to happen in the very short future anyway. So it's...

36:10

Aaron van Wirdum Well, you also mentioned you just call it a recommendation, and I would agree with that in the sense that a default is already a recommendation.

36:17

Sjors Provoost Yeah, as long as everybody sticks to the recommendation, then nothing changes. But that also goes to the, you know, what is the reality now? How many people are going to turn on this setting? I don't know. But then you have people like Peter Todd who are now trying to accelerate this new reality anyway by... Running a bounty program I don't think we talked about that yet.

36:36

Aaron van Wirdum Yeah let's let's mention that as sort of last part of this episode because I've seen something about it but I haven't kept up with that thread. Like that was another side argument somewhere happening somewhere where I wasn't paying attention… but there was a bounty program by Peter Todd is that right?

36:53

Sjors Provoost So one way to get to avoid the problem of having two different settings or you know two different mempools or or having this debate go on forever is to just basically make miners use full RBF at such a scale that it just becomes even more unsafe to keep the opt-in RBF version around. And so that's what he tried to do. I think he created an address where people can donate, but also just put some of his own money in. And he's basically making transactions that don't signal RBF and pay like one Satoshi per byte. And then he puts another transaction on top of it that pays a hundred dollars in fees and then basically any miner that doesn't run full RBF would be stupid in this case because they would miss out on 100 bucks. And then if you do that at the right time, so I guess when fees are very low but the mempool is full. So I don't know some some moment when there's a strong incentive to to snap up that extra hundred dollars you can very cheaply create an incentive to tilt the ecosystem in one direction yeah and he can do that.

37:55

Aaron van Wirdum So and how is that going?

37:56

Sjors Provoost I don't know.

37:59

Aaron van Wirdum So I think...

38:00

Sjors Provoost I mean, today or a couple of days ago before we were recording it, Binance decided to consolidate all their transactions, which actually filled the mempool very quickly. And then it's actually already trivial to do double spends because all the low fee transactions would have been booted out of the mempool and you can simply create a brand new transaction with a higher fee and it would just propagate everywhere. So that kind of interferes with the experiment. I don't know what the stats are.

38:25

Aaron van Wirdum I thought the takeaway so far was that miners were not using full RBF. That's what I sort of saw from the corner of my eye in one of the side discussions that I didn't really follow. But I think that was the latest update on that end.

38:38

Sjors Provoost My hope is that there'll be some good dashboards that observe exactly which pools are doing that and which pools are not, which gives you some indication of what the odds are.

38:49

Aaron van Wirdum Yeah. Are you going to make that? This sounds like a job for Sjors.

38:52

Sjors Provoost No, what I do is I go to projects like miningpoolops or miningpoolobserver. I think it's miningpool.observer, the domain. And I go to the GitHub repo and I make an issue saying, hey, it'd be good to monitor this. And then somebody else gets really excited by that and does it.

39:09

Aaron van Wirdum Smart, smart. That's how smart people work.

39:11

Sjors Provoost Yep.

39:13

Aaron van Wirdum Okay, Sjors, I think that's it for me. Did we forget anything? Do you want to add anything? Do you want to tell our listeners something before we end the episode?

39:22

Sjors Provoost No, I believe Bitcoin Core version 24.0 is coming very soon, TM.

39:27

Aaron van Wirdum Including the option to switch on full RBF. That was not removed in the end, right?

39:34

Sjors Provoost Exactly. And you should download that new release anyway and try it out. And so thank you for listening to Bitcoin.

39:40

Aaron van Wirdum Explained.