比特币的力量与Gloria Zhao - WBD810

在深入理解代码和加强代码方面，我们还有很多工作要做。有时需要进行侵入性修改才能提高可扩展性。这并不是说永远不要改动代码，而是要谨慎行事，避免盲目追求速度而导致问题。

There's just so much to do just in terms of understanding the code better and fortifying it. Sometimes you need to invasively change something in order to make it more scalable. So it's not don't ever change the code, you know, be be conservative and not try to move fast and break things.

你好。最近怎么样？周三快乐。欢迎收听《比特币做了什么》播客，本期由铁巨人（原名为Iris Energy）赞助播出。铁巨人正利用其新一代数据中心，以100%可再生能源推动比特币挖矿和人工智能的未来发展。

Hello there. How are doing? Happy Wednesday. Welcome to the What Bitcoin Did podcast, which is brought to you by the massive legends of iron, formerly known as Iris Energy. Iron is using their next generation data centers to power the future of bitcoin mining and AI using 100% renewable energy.

铁巨人仍是同一家企业，目标不变，只是换了名称。我是主持人Peter McCormack。今天我们有幸邀请到出色的Gloria Zhao参与节目。去年年底时，我很幸运能在非洲与Gloria共度时光，之后我们一直努力安排她上节目。

Iron remains the same business with the same goals now with just a different name. I'm your host, Peter McCormack. And today, we have the amazing Gloria Zhao on the show. Now at the end of last year, I was lucky enough to spend some time with Gloria out in Africa. And since then, we've been trying to get on the show.

Gloria是比特币核心维护者和开发者。本次访谈我们将探讨持续升级协议的重要性（而非僵化）、告别谷歌迎接比特币序数和内存池垃圾信息、以及比特币核心维护者的力量。与Laura的对话非常愉快，她令人惊叹。很高兴能深入探讨这些话题。

Now Gloria is a Bitcoin core maintainer and developer. In this interview, we get into the importance of continually upgrading the protocol as opposed to ossification, ditching Google to welcome Bitcoin ordinals and mempool spam, and the power of Bitcoin core maintainers. I really enjoyed sitting down with Laura. She's amazing. Enjoyed getting into this.

希望你们一如既往喜欢本期内容。如有问题或反馈，请联系我：hello@whatbitcoindid.com。一、二、三。

Hope you liked it as well as ever. If you have any questions or feedback, hit me up. It's hello at what bitcoin did dot com. One, two, three.

什么？哦，我该说3、4、5、6吗？

What? Oh, do I say 3 456?

没错。现在节目正式开始。这是倒计时。是的。

Yeah. That's the start of the show now. That's counting. Yeah.

我懂数字。烦死你。我知道很多数字。

I know numbers. Annoy you. I know a lot of numbers.

你应该的。你是你是比特币界最重要的极客之一。

You should do. You're you're one of the most important nerds in Bitcoin.

我的观点是我其实没那么重要。真的吗？我总是否认那种认为我拥有某种权力的说法。好像我能影响比特币似的。哇哦。

I my argument is that I'm not actually that important. Really? The thing that I always push back on is the idea that I have some kind of power. Like I have some ability to influence Bitcoin. Wow.

我们已经开始了，对吧？

We're we've gotten started already, haven't we?

是啊。我们直接进入深水区了。好吧。你个人没有那种权力，因为有制衡机制。

Yeah. We went right off the deep end. Okay. You don't have the individual power because of the checks and balances.

因为我能做的每件事都是完全公开透明的，所有人都看得见。而且除了编写代码外，我什么都做不了——那些代码可能被运行，也可能不被运行，还要经过一段时间我们才会发布新版本。总之我们会深入讨论这个的。

Well, because there's nothing that I can't do there's nothing I can do that isn't fully public that everybody can see. And there's nothing I can do, like, other than create code that may or may not be run by somebody after some amount of time after which we make a release. Anyway, we'll we'll get into this.

那那我们就来铺垫一下。给听众们解释解释你是做什么的

So so let's let's set it up. Explain explain to people listening what you do

好的。

Okay.

或者你是什么身份。

Or what you are.

好的。我直接开始自我介绍吗？

Okay. Should I just intro?

我们开始了吗？五分钟前就已经开始了。

Have we started? We started we started five minutes ago.

好的。大家好，我是格洛丽亚，从事比特币核心开发工作已有四年多。

Okay. Okay. So hi, I'm Gloria. I work on Bitcoin Core. I've been doing that for a bit more than four years.

我担任所谓的维护者已近两年，这仅意味着我有权限将代码推送到托管比特币核心的代码库。

And I've been what's called a maintainer for almost two now. And that just it means nothing other than I'm able to push code to the the repository in which Bitcoin Core is is hosted.

实时代码库。

The live repository.

是的。我们有一个GitHub仓库，里面存放着Bitcoin Core的所有源代码，它的开发方式与许多其他开源项目类似。通常开源项目只有少数人拥有写入权限，可以推送代码。除此之外，我们还设置了一个脚本，用于检查所有推送的提交和差异是否附有PGP签名。

Yes. Yes. We have a GitHub where we have a repository of all the source code in Bitcoin Core, and it's developed much like many other open source projects. And typically open source projects have a small number of people who are able to, have write access so they can push code to it. And we have, you know, in addition to kind of the GitHub permissions, we have a script that checks that there is a PGP signature on all of the commits that are pushed, all the diff that is pushed.

我记得当时跟你说这些时，你的反应是‘哦，我该怎么查证？’其实有个公开文件叫‘trusted keys’，里面列出了五位成员的PGP密钥，只有这些人的签名才能被接受用于推送到仓库的提交。

And so this is all I remember when I said this to you, you're like, oh, how do I find out? There is a public file called trusted keys, and it lists PGP keys of the five people who whose signatures are accepted for these commits that are pushed to the to repository.

所以你是这五人之一？

So you're one of the five?

是的。不过我还想说，我们有很多依赖项。人们通常只关注Bitcoin Core仓库，却很少提及libsecp的维护者——那些能向密码学库推送代码的人，而密码学库在比特币验证过程中生成和验证签名是非常关键的部分。

Yes. However, another thing that I say is there we have all these dependencies. Right? So people, they often focus on like the Bitcoin Core repository, but they don't talk about like the libsecp maintainers. So the people who can push code to the cryptography library, which is a very, very important part of how signatures are created and verified within Bitcoin validation.

以上就是全部了，但我觉得这些讨论没有意义，因为我的观点是：虽然推送代码确实带有某种特权，但我们的所有操作都是完全公开可验证的。从我们推送代码到节点运营商最终在网络上运行，再到可能被广泛采用，中间有太多环节了。这就是我的论点。

That's all but I think all this is moot because my argument is, yes, I I I recognize that there is like an amount of, I guess, privileges with being able to push code, of course. But there is nothing we can do that isn't fully publicly verifiable. And like there are so many steps between us pushing code and it being eventually run on the network by a node operator and maybe like becoming widely used thing on the network. So, like, that's my argument.

你看过新版《星球大战》里绝地议会那段吗？

Have you you know the new Star Wars where the Council of the Jedis? Have you ever seen that?

没看过，抱歉。

No. I'm sorry.

我...那个...我待会儿会给你看一张照片。我把你和维护者们想象成绝地武士团的样子。嗯。就坐在那里，守护着我们的比特币网络。

I that's that's what I I'm gonna show you a photo afterwards. I think of you and the maintainers as like the Council of Jedis Mhmm. Sat there, roarding over our Bitcoin network.

所以...好吧。我觉得我们对代码理解得很透彻。是的。我们可能花了很多时间思考安全问题，也许我们知道哪些漏洞被发现以及它们是如何被修复的。但...所有这些其实都是公开的。

So I okay. So I I think we understand the code very well. Yeah. And we are maybe we've spent a lot of time thinking about the security issues and maybe we know things that about what has what bugs have been found and how they've been patched. But but like, all this is public.

好的。但你能单独推送代码上线吗？还是需要你们五个人都同意才行？

Okay. Can but can you push something live on your own or or do all five of you have to agree?

我们不是以委员会的形式运作的。

So we it's we don't do it as like a committee style.

好的。

Okay.

是的，我可以在不征求其他人同意的情况下推送代码。不过...假设我推送了所有人都反对的内容，结果会怎样呢？它基本上会立即被回滚。而且...取决于情况有多严重，他们甚至可能撤销我的密钥。这些通常都会在软件正式发布供公众下载前很久就处理完毕。

So, yes, I can push without, like asking the other ones for permission. However, like we can play out the scenario in which I push something that everybody disagrees with. What would happen is it would be reverted pretty much immediately. And I I mean, depending on how crazy it was, like, they can also remove my key. And this would all happen usually a very long time before the software is actually released and put out in the public for people to download.

当然代码就在那里，如果有人想下载还是可以的。但过去确实发生过维护者做出非常疯狂举动的情况...很久以前加文就遇到过这种事，对吧？

Of course, it's there and if someone wants to download it, they can. But like what has happened in the past is we've had situations where a maintainer has done something very crazy and they they were it was Gavin a long time ago. Right?

是的，当然。

Yes. Of course.

没错。人们总是担心，比如，如果维护者单方面合并了一个共识变更之类的会怎样？但这并不意味着软分叉会立即在网络上线。对吧？从代码提交、发布、到网络中的个体下载，再到激活，中间有太多步骤了。

Yeah. And so like, people are always worried about like, oh, what happens if a maintainer unilaterally merges like a consensus change or something? That doesn't mean that the soft fork has gone live on the network immediately. Right? There are so many steps between code being pushed, code being released, code being downloaded by individuals in the network, then becoming active, you know.

对。所以回答你的问题，是的，我可以。但就像你说的，有制衡机制或诸多原因，这并不意味着我拥有任何单方面的权力。

And yeah. So to answer your question, yes, I can. But like there are, like you said, checks and balances or many reasons why that doesn't mean that I have any kind of unilateral power or anything.

所以，如果你偷偷摸摸地，而绝地委员会的其他成员都在睡觉，你推送了一些疯狂的东西上线

So so if you if you were being sneaky and the rest of the council of the Jedi's were all asleep and you push something crazy live

嗯。

Mhmm.

还有很多人会检查代码，验证它。通常，这只会演变成社交媒体上的闹剧：'格洛丽亚疯了，她干了这种疯狂事'。

There's plenty of other people out there who'll be checking the code and Yeah. Verifying it. And then, typically, this would just be this would become some kind of drama on social media. Gloria's gone crazy. She's done this crazy shit.

是的。然后人们就不会下载它，他们会说：'别下载这个'。

Yeah. And then people wouldn't download it. They're like, don't download it.

是的。而且这种情况过去已经发生过，对吧？所以，我不是在凭空想象什么幻想世界。你知道，我们有一个非常强大的开源社区来验证一切。

Yeah. And this has has has happened in the past. Right? So, like, I'm not just, like, theorizing, like, this fantasy world. Like, you know, we have this very robust open source kind of community that that verifies everything.

这就是我的观点。什么

So that's that's my point. What

你学过计算机科学吗？

did did you study computer science?

学过。是的，我在伯克利学的计算机科学，那段经历很棒，我非常喜欢。我认为那是个很好的基础。

I did. Yeah. I went to school for computer science in Berkeley, and it was it was great. I I loved it there. And I think it was a great foundation.

而且我就是在那儿接触到比特币的。所以我不确定

And that's where I found Bitcoin. So I don't know if

等等。你之前在一次...（口误）你告诉过我这件事。对，你跟我提过和你父母的一次对话。

Hold on. You told me about this when we were in a lie. Yeah. You told me about a conversation with your parents.

哦，对。我想想...我一直在做这个...这是什么时候的事？大概2020年左右吧，那时候资助项目还比较新，Brink都还没成立。

Oh, yeah. Yeah. I I think so I've been doing this what was this? This was in like 2020 or so, where I think kind of the grant programs were kind of new. Brink didn't exist yet.

我是Brink的第一位受资助者，Brink是一家非营利组织，它从比特币社区筹集捐款并资助比特币核心开发者。这个组织基本上是在那一年成立的。我想你采访过John对吧？关于这件事的。但基本上，当我刚开始时，开源领域的情况还比较常规。

So I was the first grantee at Brink, which is a nonprofit that that collects donations from the Bitcoin community and gives them to Bitcoin core developers. That, you know, was founded basically that year. And I think you interviewed John Yeah. For that. But basically, when I first got started, it was it was kind of more status quo what tends to happen with open source.

互联网和我们日常使用的许多东西都建立在开源软件之上，这些软件是有人免费编写后以某种开源许可证发布在网上的。他们通常会说，随便用吧。很多时候，庞大的系统建立在开源软件上，而代码作者并未获得报酬，他们可能仍在维护，也可能没有。很多时候他们确实在维护，为了支持用户需求的功能、推送安全补丁、更新依赖项以适应操作系统或编译器的变化，作为下游项目必须跟上这些更新。

So a lot of the Internet and a lot of the things that we use on a day to day basis are built on open source software, which like someone wrote for free and then put it on the internet under some open source license. They're like, yeah, you're free to use it, whatever. And, you know, often there's like huge things built on open source, software that the person who wrote like the person who wrote the code was not is not paid for. And they may or may not still be maintaining it. A lot of times they are still maintaining and they're doing all this work to kind of support the features that users have asked for to, you know, push in like security fixes, to like update for their dependencies, like, you know, with operating system changes or compiler changes or library, you know, updates, you also have to like keep up with those updates, as a downstream project.

但往往没人为此付费，我想这是因为我们的文化并不重视这一点。通常我们等待中间商介入，剥削生产者并从消费者那里榨取租金。抱歉这么说，但过去捐赠模式并不太可持续。然而对于比特币核心这样的开源协议，这几乎是唯一的方式。

And oftentimes that just like nobody pays for because, you know, I guess our culture is not really, about it's we usually, I guess this is like a social commentary thing. Like, we usually wait for an intermediary to step in and like exploit the producer and extract rent from the the consumer. And sorry to to say that. But, you know, typically, the donation model hasn't been very sustainable in the past. But that's kind of the only the only way to do it with like Bitcoin Core, like Bitcoin open source protocol stuff.

但你本可以去谷歌，拿50万美元年薪，还有股票期权，开崭新的特斯拉。

But you could have gone to Google, got a half a million dollar salary, share options, had a brand new Tesla.

是的。差不多在我接触比特币的同时，我正在申请实习。我确实去了谷歌实习，本来还打算去Facebook，但我反悔了。当时觉得挺酷的，不过...

Yeah. So around the same time I found Bitcoin, I I, you know, I was applying to internships. So, yeah, I ended up I ended up doing an internship with Google and I was gonna do one with with Facebook, but I reneged on that. And I thought it was so cool. But I yeah.

必须说明，我在谷歌度过了非常棒的时光，学到了很多——云服务、微服务架构、Kubernetes、Terraform等等。但比特币完美地平衡了两点：它试图解决我关心的世界性问题，同时技术层面也极其有趣。

I had a fantastic time at Google to be to be clear. I've learned a lot. I learned about like cloud, microservices architecture, Kubernetes, Terraform, whatever. And it was a great time. But like Bitcoin was like this perfect, like, balance between like something that was trying to solve a problem in the world that I cared about, and also very, very technologically interesting.

我记得刚开始研究比特币核心的头几周，我们在讨论线程安全问题，因为它是多线程的。我当时想，这我在课堂上学过。大概那时Peter Willett还写了个叫mini sketch的数学库，我的抽象代数课本知识居然派上了用场。

So I think like within the first few weeks that I was looking at like Bitcoin Core, I don't know. We were it was it was like, oh, there's like this, you know, we we're thinking about thread safety because it's a multi threaded, thing. And I was like, oh, I learned about that in class. Or, I think around that time, Peter, Peter Willett had written this like mini sketch library, which is like this fancy math. And I was like, oh, my abstract algebra class like textbook like comes in handy.

这就像是我自从上完那门课后第一次打开它。你知道，里面有非常酷的东西，比如密码学、分布式系统之类的。彼得·威尔就是个很好的例子，还有我共事的许多人，他们写代码的时间和我活的时间一样长，或者你知道，那些顶尖的计算机科学家和工程师，至少是我见过的在这个领域最高水平的人，他们正在研究这个很酷的开源项目——那种抗审查的货币体系，它正在解决世界上的实际问题。我个人认为这对软件工程师来说是个超级被低估的项目，应该有更多人觉得它比谷歌之类的工作更有吸引力。这是我个人的决定。

It's like the first time I opened it since like I took the class. And, you know, there's really cool, like, cryptography, distributed systems, like all these things. And and like, Peter Will is a great example and a lot of other people that I work with who have been coding for as long as I've been alive or, you know, these amazing, like, computer scientists and engineers kind of at the at least the highest level of this craft that I've seen working on this, like, cool open source, like, censorship resistant money thing that is solving a real problem for the world. I I personally think that it is a super underrated project for software engineers that that more like, more people should consider it something more attractive to work on than, like, Google. I mean, that was a personal decision of mine.

我父母...抱歉。我刚要讲我父母当时的故事，他们问‘你为什么要做这个？你说你要做这个是什么意思？你还在从人权基金会拿资助。’

And my parent yeah. Sorry. Was telling you the story where my parents were like, why why would you do this? What do you mean you're gonna work on this? You're getting a grant from Human Rights Foundation.

‘这和那个有什么关系？’你懂吗？

Like what does that have to do with that, you know?

‘教育要花这么多钱？’

Like education cost this much?

是的。但你知道，换个角度看，我特别幸运能接受这么好的计算机科学教育。我的父母支持这个决定，现在我能开发这么酷的软件，真的觉得自己非常幸运。

Yeah. Yeah. But, you know, like I'm, you know, that's another angle to it is like, I'm so I'm so privileged that I got this amazing computer science education. I have parents that were supportive of this decision, and now I get to work on this really cool piece of software, and I feel really really lucky to be able to do that.

他们现在理解了吗？

Do they understand it now?

我用过那种翻页板。有时回家看到车库里的翻页板上写着‘比特币入门’，我就觉得‘嗯，他们现在懂一点了’。我妈现在甚至会交易退休账户里的比特币ETF。挺酷的。

I've I've done like the flip chart. Sometimes when I go home and I see like the flip charts that I was in the garage where it's like intro to Bitcoin, I'm like, yeah. They they get it a bit more now. My mom, who like trades their their retirement account is like buying Bitcoin ETF. Cool.

我是说，一开始你会觉得，妈妈，你并没有真正持有资产。你可能会被坑。

I mean, to start, you're like, but mom, you're not holding the actual asset. You could be rugged.

是的。我是说，我对这事有些看法，我觉得...对吧？从意识形态上讲，我们为什么要参与这个，对吧？意识形态上，自我托管就是关键所在。对吧？

Yeah. I mean, so I have some opinions about this where I think, yeah. I think like ideologically why why we're getting into this, aren't we? Ideologically, yeah, self custody is like, that's the point. Right?

不过我认为普通人在自我托管方面可能不太擅长

I think the average person though is probably less good at self custody

如果你是指我父亲。因为我，他对比特币产生了很大兴趣，但他每天就只会查价格，然后说'涨了'、'跌了'。要是让他尝试自我托管比特币，他肯定会搞丢的。

than If you meant my dad. My dad's got so much interest in Bitcoin because of me, but all he does is check the price every day and he's like, the p is up, p is down. Yeah. If he tried to buy in self custody Bitcoin, he'd lose that. Yeah.

没戏。

No chance.

是的。所以我真的非常犹豫是否要向人们推荐那些需要相当技术素养和近乎偏执的操作。懂吗？这不是说我在支持使用那些可能靠谱也可能不靠谱的交易所。但我会保持公正。

Yeah. So I'm really I'm really hesitant to, like, make recommendations to people that require quite a bit of technological literacy and like paranoia. You know? So this is not to say that I'm endorsing like, you know, using these exchanges that are that may or may not be reliable. But like, I'll be fair.

我...我理解了吗？是的。我不想建议人们去做有风险的事，尤其是我妈妈。

I I like You get it? Yeah. I'm not like I'm not trying to advise people to do things that are risky, especially my mom.

比特币在计算机工程领域有多独特？

How unique is is Bitcoin in terms of computer engineering?

哦对。我前几天刚向一群软件工程师解释过这个，我可能会换个说法。但确实，比特币有个非常特别的安全模型。它的设计初衷就是在存在恶意攻击者的情况下仍能正常运行。

Oh, yeah. I mean, so I was just explaining this the other day to a lot of software engineers, I guess I would word it differently. But there's yeah. There's this crazy security model. So Bitcoin is kind of designed to work in the presence of adversaries.

对吧？

Right?

是的。

Yeah.

就像推特有自己的术语来描述这种状态，这是个难题。比特币网络应该能够持续产生区块、处理支付等，即使遭受攻击。我们编写的软件追求去中心化，对吧？这是意识形态层面的——网络的价值就在于它是由众多不同实体和组织共同运行的分布式网络。

Like Twitter has their own lingo to describe this nation, you know, it's a hard, you know. And and it's supposed to be able to, like, continue to produce blocks, continue to process payments, etcetera, even if people are attacking it. And the software that we write is supposed to be like, we're trying to go for decentralization. Right? That's like a ideological like, that's how the network brings the value that that it has is that it's a decentralized network that is run by lots of different entities and organizations.

对吧？但从实际编码角度来看，这意味着我写的代码要尽可能让所有人都能使用。我们支持多种平台——Windows、Mac、Linux等等。在计算复杂度方面，我们不想只编写能在AWS服务器或高性能设备上运行的代码。

Right? But what that looks like in practice from a code writing standpoint means that, like, I write code that it is hopefully usable by anybody. Right? So like we support like so many platforms, Windows, Mac, Linux, whatever. We, in terms of computational complexity, like we don't want to only write code that could be run on like AWS servers or like huge beefy thing.

它应该能作为后台任务在你的笔记本上运行。很多人就是这样运行全节点的，甚至应该在树莓派上也能运行。我认为这是个很好的基准——你可以在小小的树莓派上运行全节点：转发交易、估算手续费、为你处理所有这些事务，很多比特币用户都这么做。但这两者（性能和普适性）确实很难平衡。

Like it should be runnable as like a background task on your laptop. A lot of people run their full nodes that way. It should work on a Raspberry Pi. I think that's a very good benchmark where you can run a full node that relays transactions, can do fee estimation, can, you know, do all these like things for you on this tiny little Raspberry Pi box, which a lot of people who are Bitcoin users do. But those two things are really difficult to balance.

所以我们处于一个非常狭窄的操作空间里，理论上任何人都能运行这个系统。但当你在笔记本电脑或其他设备上启动比特币D时，它所做的就是连接到网络上的随机节点。按照设计，我们并不知道这些节点的身份。对吧？我们希望它是一个匿名或伪匿名的网络。

And so like we're in this very narrow space of operation where, yeah, anyone should be able to run this. But then when you turn on Bitcoin D on your laptop or whatever, what it's doing is it's connecting to randos on the network. We don't know what their identity is by design. Right? We want we want it to be an anonymous or pseudonymous network.

它会尝试下载数据，寻找主链，下载区块或发现新交易，以及寻找要连接的节点地址等等。这意味着我们必须防范随机攻击者连接到节点并发送垃圾数据，比如可能导致你的节点崩溃、停滞或耗尽内存。这些都属于我们认定的最高严重级别的攻击。

And it like tries to download data. It tries to find the main chain. It tries to download blocks or like find out about new transactions or addresses of nodes to connect to, etcetera. And that's that's what that means is we have to plan for like a random attacker connecting to nodes and sending you like garbage data that like, for example, can crash your node or cause your node to stall or cause your node to run out of memory. And these are all attacks that we would consider like the high one of the highest severity.

我提到这个是因为人们总是问我：还有什么需要做的？或者：你在忙什么？代码不是已经完成了吗？

And I bring this up because basically, people always ask me like, what is there still to do? Or like, what do you work on? Like, isn't the code finished?

直接伪造它就行。

Just falsify it.

是的。我试图传达这样的信息：有很多事情可能会出错。人们会说51%攻击在经济上不可行，从公钥推导私钥是不可能的。

Yeah. Yeah. And I tried to give this message of like, there's a lot of like, you know, things that can go wrong. Right? Like people are like, but 51% attacks are not economically viable and it's impossible to get a private key from a public key.

就像，格洛丽亚，你没看安德烈亚斯·安东波罗斯那个解释比特币安全性的视频吗？我说：是的我同意。但我讨论的是，在我们设想的比特币未来里——希望它能被数百万人使用——存在无数可能破坏这种愿景的因素，包括我刚才提到的网络攻击。

Like, Gloria, you didn't watch this Andrea Santonopoulos video where he explained why Bitcoin is secure. I was like, yes, I agree. I agree. Like, I'm talking about like, there are a lot of there's like a million ways that like we all envision like this future for Bitcoin where we want it to be used by like millions of people and all these things. There are a lot of things that can undermine that, including for example, the networking attacks that I was talking about.

如果有人能崩溃节点，比如矿工能让其他矿工崩溃从而在后续几个区块获得先发优势。或者考虑区块审查攻击：你生产了一个区块，如果有人能阻止其他所有人听到或下载这个区块，这就是非常严重的问题。还有日蚀攻击——整个点对点结构的核心就是每个人都能自由连接，理想情况下能找到能告知主链信息的优质节点。如果这个机制被破坏，攻击者能让你连接的所有节点都变成他们控制的，导致你听不到主链或竞争矿工的区块信息。这些攻击都不需要51%算力就能实施。

So if somebody is able to crash nodes or like for a miner, for example, is able to crash the other miners and they get head start on like the next few blocks. Or let's say we would care a lot about block censorship attacks. So like you produce a block, if somebody can prevent everybody else from hearing about that block or downloading up that block from you, that is a that is a really serious problem. Or if you've heard of Eclipse attacks where, you know, again, like the whole point is we have this peer to peer structure where everyone can talk to whoever and like hopefully find a good set of peers that is able to tell them about the main chain. If that is compromised and like somebody's able to eclipse you as in like be everyone that you're connected to and you don't hear about the main chain or you don't hear about the competing miners like blocks, these are all really like I mean, they're not the same thing as well, I mean, if a miner is able to do this, that's like, you don't need 51% hash rate in order to attack these people.

对吧？就像你直接屏蔽他们，让他们无法得知其他哈希率的信息。

Right? Like, you just block them from hearing about all the other hash rates.

但这些算是攻击吗？

But are these attacks?

这些都是理论上的攻击，只是澄清一下。

These are theoretical attacks, just just to be clear.

那么如果这是理论上的攻击，你们会尝试实施这种攻击以便知道如何防御吗？

So and so if if it's a theoretical attack, do you try and attempt to do the attack so you know how to defend against it?

实际上，我们通常做的是分析代码并编写测试。我们试图隔离区块下载逻辑，尽可能多地测试各种可能发生的场景。我们有很多监控节点，如果观察到有人试图发送变异区块，或者有人通过先发制人发送垃圾信息来阻碍区块下载，当我们要求他们提供正确信息时却故意拖延。既有理论上的攻击，也会观察到有人在做奇怪的事情。

Well, in practice, what we're often doing is we're, like, analyzing the code and we're writing tests. We're like, we're trying to isolate the block download logic to, just throw random stuff, like, just, like, test as much as possible, like all the scenarios that we can possibly happen. And, you know, we have lots of monitoring nodes where we're seeing like, you know, if we observe, oh, wow, there is actually someone trying to send us mutated blocks or there is someone trying to stall block download by like, being the first guy and then like sending us garbage. And then like when we ask them for, you know, the correct information, they just like, you know, stall. Like, there are like theoretical attacks and then there's also like, we'll observe people doing weird stuff.

这有时会促使我们去调整那些参数，或者意识到可能需要重新设计某些部分以做出更好的权衡。抱歉我有点跑题了。

And that will sometimes lead us to go and tweak those parameters or be like, oh, yeah, maybe we should, you know, we we should redesign this a little bit so that, we're making a better trade off. So for like sorry if I'm rambling by the way.

没关系。

You're good.

所以关于交易中继，我思考了很多。几年前有个案例是交易停滞问题，比如说有个交易正在广播，大多数人都知道这事。我不想多次下载那个交易，只需要下载一次就够了。

So with, I I think a lot about transaction relay. There was one a couple years ago that was, transaction stalling where, let's say, you know, there's a transaction being broadcast. Most people know about it. I don't wanna download that transaction multiple times. I only need to download it once.

对吧？所以我们有这类节省带宽的措施。重申一下，节省带宽很重要，因为不是每个人都有超高速WiFi，对吧？我们希望运行节点是容易实现的。所以通过只下载一次来节省带宽。

Right? So we have these kind of bandwidth saving things. Again, bandwidth saving is important because not everybody has extremely high WiFi speeds, right? We want it to be accessible to run a node. So we're trying to save bandwidth by only downloading once.

因此我们每次只会向一个节点请求发送该交易。但如果第一个节点——我是说第一个对等节点——没有发给我们，我们就应该重试。对吧？而且我们不应该只试两次，因为要避免出现审查漏洞——比如某个节点心想'好啊，我就当你前三次请求的交易来源'，然后突然你就收不到这个交易了。

So we're only gonna ask one peer at a time to send us that transaction. However, if we try the first person, the first peer, I mean, and they don't send it to us, we should try again. Right? Or we should like and we shouldn't only try two times, for example, because we don't want there to be a censorship vector where, you know, a peer is like, okay, I'll be the three the first three people that you asked for this transaction from. And then all of a sudden you're not going to hear about this transaction.

那就会形成真正的交易审查问题。所以后来基本上是对交易请求机制进行了重新考量，通过模块化改造和修复这些问题使其更健壮、更易于理解，还添加了大量测试用例来验证——比如'在所有这类场景下我们最终都能下载到那个交易'。我们最近大量使用模糊测试，不知道你听说过没？要我详细说说吗？

That would be a transaction really censorship problem. And so, yeah, there was basically, there was kind of a TX request kind of reconsideration kind of there was like modularization plus fixing up some of those things to make it more robust and make it really easy to reason about, and then added a bunch of tests to kind of check our understanding of like, oh, all these in all these scenarios, yes, we would end up downloading that transaction. And then we have what are we've been using fuzzing a lot. I don't know if you've heard of fuzzing. Should I ramble more?

详细说说，给我讲讲模糊测试。

Ramble more. Tell me about fuzzing.

好的。通常软件工程师写测试的方式是——比如我刚描述的场景，你会搭建一个模拟节点，再连几个额外模拟节点，然后模拟攻击并断言行为是否符合预期。这种测试需要你精确设计测试用例。

So, okay. So typically the way that a software engineer might write tests is, for example, you'll write the test scenario that you that I just described. Right? Where you're like, okay, let's spin up a note, like a fake node, and then we'll spin up a few extra fake nodes to connect to it, and then we'll simulate the attack and we'll assert that, like, the correct behavior is happening. And so you have to provide exactly what the test is.

而模糊测试能自动生成测试用例。你只需要写个框架，定义一些不变量——比如每隔一段时间检查所有不变量是否成立。然后模糊测试引擎会智能地开始发送随机数据，再根据代码覆盖率等情况进行调整。

Fuzzing, automatically generates tests. So what you'll do is you'll write kind of more of a skeleton, and like, you know, it'll check invariants like, you know, after each state like after each period of time, you check that all the invariants are still true, whatever those may be. And then a fuzzing engine will kind of intelligently just like it'll it'll start by like sending random stuff. Right? And then it'll figure out based on, for example, coverage of the code.

它会查看你执行了哪些代码行。如果发现某些输入能深入到模块更内部的代码，我们就会创建更多类似的输入。这基本上类似于生成随机输入，但它具有一定智能性，随着时间的推移，它能发现你可能手动编写测试时想不到的路径。

Like it'll look at what lines of code you've hit. And if they're like, oh, these inputs like hit like a further a line of code deeper into the module. Okay. Like we're gonna create more inputs that look like that. And so it basically it's similar to like creating random inputs, but it is somewhat intelligent and it'll like over time just like find the paths that you maybe wouldn't have thought of if you were just manually writing the test.

我们现在经常使用这种方法——对所有东西进行模糊测试。过去几年里，源代码的模糊测试覆盖率大幅提升，并发现了相当严重的缺陷。

So we use this all the time. Now like we're like fuzz everything. Fuzz everything. And and like the amount of fuzz coverage over the source over the source code has increased like dramatically over the past few years. And it's found like pretty severe bugs.

在设计某些功能时它也很有帮助，比如测试计算复杂度时。我们会说'用模糊测试看看平均运行时间'，或者'看看模糊器能否生成导致长时间运行的输入'。这样我们就不只是理论分析算法的最坏情况时间复杂度，而是能实际观察'我们是否遗漏了什么？真实表现如何？能否添加逻辑来捕获这些极端情况行为？'

It's helped design certain certain things as well when when you're like trying to test the computational complexity. And you're like, okay, like let's like fuzz it and see kind of what the average runtime of it is. Or we'll fuzz it and we'll see if the fuzzer can come up with an input that causes it to spin like for a really, really long time. And we're like, okay, like instead of just theoretically analyzing what's the worst case runtime of this algorithm, like maybe have we missed something or like what does that actually look like? Or like can we add, you know, an extra piece of logic in there that like catches these really like worst case behaviors?

它既能辅助设计，又能帮助测试，模糊测试真是太棒了。

Or, you know, like, it it helps design things. It helps test things. It's fuzzing's great.

我认为大多数人（尤其像我这样的）想到比特币开发时，往往只考虑扩容问题。

I I think most people, especially like me, when we think about Bitcoin development, we tend to think of just scaling.

嗯。

Mhmm.

大家总在讨论'如何扩容？有什么新功能？'，但很少考虑有多少工作是在防范恶意攻击或优化效率上。嗯...这类工作是如何在核心开发者之间分配的呢？

How are gonna scale? What's the new what what are the new features come in? I don't think many of us tend to think about how much work is done, planning for adversarial attacks or efficiency. Mhmm. And and how how is that kind of work split up amongst core developers?

是否有专人负责攻击方向，专人负责效率优化？你们是不是在每个领域都有相应的专家？

Is it someone specifically is working on attacks, someone's on efficiencies? Do do you tend to have, like, specialists in each area?

其实并没有专门的协调或管理机制，但确实有各自专长的人。比如我要介绍下我在Brink的同事Nicholas，他是模糊测试专家，后来也自然成为了我们的安全工程师——毕竟他总能发现各种漏洞。他非常出色。另一位同事Michael Ford（你可能认识他叫Fanquake）专攻构建系统，这也让他自然承担了很多发布流程的工作。

So there's no kind of coordination or But management, of there are people who specialize. So I'm just going to show my colleague, Nicholas, who works with me at Brink. He he's the fuzzing guy, and he's kind of become our security engineer as well because naturally, he becomes the guy that finds all the bugs. And and he he's great. I my other colleague, Michael Ford, who I think you've met or Fanquake, he specialized in build system stuff, and that also naturally leads him to do a lot of the release process stuff.

我们也可以聊聊发布流程，我觉得这部分特别有意思。大家会叫我内存池专家——当然不止我一个人负责这块，但我确实在这个领域投入了大量时间。通常如果有人提交了内存池相关的PR让我评审，我都会处理。

And we can talk about the releases release process as well because I find it very interesting. And then people will call me like the Mempool person. I'm not the only person that works in Mempool, definitely. But like, you know, I spent a lot of time in this area. And typically, if somebody, you know, opens a Mempool PR and you ask me to review it, I will.

对吧？Ava Chow主攻钱包开发。不过我觉得大家其实都会涉猎多个领域，只专注一件事反而会很无聊。但确实会自然形成'这个人特别擅长这个'的情况。

Right? Ava Chow, she works on wallet. But I mean, everyone does a bit of everything, I think. And it would be very boring to spend all of your time doing one thing. But I think there are you know, we've we've naturally been like, okay, this person is really good at this.

而且他们拥有最专业的知识。我认为我们的工作职责之一就是要了解整个系统如何运作——这非常重要，因为所有东西都如此复杂且高度关注安全性。

And, you know, has the most expertise. And I think part of our job is to just know how stuff works. I think it's really important because everything is so complex and it's so, like, security oriented.

但最神奇的是，在没有协调的情况下所有事情都能推进。你懂我意思吗？比特币就是这样运作的：总有人在持续升级、修复漏洞、开发新功能。而这种去中心化、无协调的运作方式居然能成功。

But it's so interesting that without coordination, everything just kinda gets done. You you know what I mean? Like but but but what I'm saying is Bitcoin works. There are people constantly upgrading, fixing things, patching things, working on new ideas. The fact that it happens because in a decentralized, uncoordinated way.

因为想想看，虽然我们做的是软件业务，但谷歌绝不可能允许这种模式——'随便选个想做的项目，准备好了就直接上传代码'。他们必须进行完全协调的管控。

Because if you think of, yes, I'm working on a software business, but I'm pretty sure Google wouldn't have it there. Oh, you just just pick what you wanna work on and, you know, when you're ready, just upload it and, you know, we'll check-in. They would be entirely coordinated.

是的。是的。我想说的是，很多参与比特币核心开发的人都非常无私，想做出贡献。我同意确实存在一种现象，就是大家想做啥就做啥，结果没人愿意做代码审查，没人愿意干那些枯燥的工作。

Yeah. Yeah. I I think I I I would just say that a lot of the people that work on Bitcoin Core are so, like, altruistic and want to be useful. And, like, I agree there is a little bit of like, oh, you do whatever you want and therefore like nobody wants to review back like nobody wants to do the boring stuff.

发布那些无聊的东西。比如

Post the boring stuff. Like

发布流程。我之前提到过反向移植，不如就详细说说。作为维护工作的一部分，我们发布一个版本后，会持续维护两年左右。大概每半年——抱歉——基本上每半年我们会发布一个新的主要软件版本。

release process. Or so I guess one I I said backport, so I might as well get into it. So as part of maintenance, right, we'll release something and then that has like two years before we stop maintaining it. So every sorry. Every six months, more more or less, we'll release a new major version of the software.

新版本会包含新功能，比如v2传输协议、数据包中继之类的。然后半年后再发布新版。但之前发布的版本我们仍会推送安全更新，所以会发布次要版本更新。

And that'll have new features like v two transport or, you know, package relay or whatever. And then six months later, we'll release a new one. Right? But the one that we released before, we're still pushing security updates too. So we'll we'll push out minor minor releases.

基本上任何漏洞修复都会先合并到主分支或开发分支，然后我们将其反向移植到当前发布的版本分支，再发布次要版本。总之，这种在不同分支上审查相同内容的工作极其枯燥，大多数人都不愿意做。而且选择哪些漏洞修复需要反向移植的过程也很费功夫，因为你必须全面掌握所有正在合并的代码——面对数百个PR，你得判断'这个会影响旧版本'，然后把它精选到旧分支。没人爱干这个。

And so basically anything that like usually a bug fix will get merged into master or development branch and then we'll back port that to the, the branch that contains a release that's currently out. And then we'll cut another minor release. Anyway, this this thing where you're just reviewing the same shit but on like a different branch is incredibly boring and most people don't do it. And the the process of picking that, the process of choosing which bug fixes to back port as well is like something that takes a lot of work because you basically have to have a bird's eye view of every single thing that's happening like in that's being merged to Matt. Like like these hundreds of PRs, you have to decide, oh, yeah.

最终效果就是：我们之前发布的版本依然能正常运行。

That's something that would affect like a previous release. And then, you know, all you do is like you you cherry pick it onto onto the the older branch and like nobody wants to do this. And and the end effect is like, yeah, that thing that we released before, it still works.

等等。所以每个分支代表一个版本？对。也就是说如果你不更新节点软件

Hold on. So is each branch a version? Yes. Yeah. And so if you're if you're not up updating your Node software

嗯。

Mhmm.

最终你会遇到一些不被支持的情况。

You eventually have things that aren't supported.

是的，对。

Yes. Yeah.

因为那个疯狂的罗马尼亚人是谁？他拒绝接受比特币最初版本之外的任何更新。

Because what is that crazy Romanian dude who refused to go anything beyond the original version one of Bitcoin?

所以我想我听说过类似的情况，有些人还在运行0.16之类的旧版本。这些人理论上认为，既然这是软分叉，我就不需要更新到新规则，应该还能继续使用——这对我们来说在意识形态上非常重要。但不幸的是，比特币核心的共识引擎与其他组件是紧密耦合的。你无法既要现代化的钱包界面等功能，又不想升级共识引擎。这种耦合性确实不好，为此我们多年来一直在推进一个项目，试图将所谓的'内核'或共识引擎模块化。这样你就能在不改变内核的情况下，获得新的GUI功能、钱包功能或其他非协议层面的前端更新。

So I I think I heard something like and I think there's people who run like 0.16 or whatever, and it's like people people who say like, well, let's say theoretically, right, like it's a soft fork, so I don't need to update to the new rules and I should still be able to use, you know, that's a very that's an ideologically a very important thing to us. And I would say, unfortunately, the consensus engine is intertwined with all the other pieces of of Bitcoin Core, like what it does. And so you aren't really able to be like, I want the modern version with, you know, the modern wallet UI and all those things, but I don't want to upgrade my consensus engine. I think it is not good that these are kind of intertwined and we have a project that we've been working on for years and years to kind of modularize the, what we call the kernel or the consensus engine. So that, like for example, you could get the new GUI feature, the new wallet features, or the new kind of more front facing, like not consensus, you know, not protocol changes without, you know, changing kernel.

所以我的观点是：不要运行无人维护的软件。这些软件包含已知漏洞，有些可能已公开披露。比如公开资料显示，有人可能发送特定数据导致你内存耗尽或节点崩溃。我强烈建议不要使用无人维护的软件。

So my point is don't run unmaintained software. It contains bugs that we do know about and have may or may not have been disclosed already. So, you know, there there are like it's public knowledge that there like, someone can send you x and then you might run out of memory or someone might crash your node or something. I very much do not recommend running unmaintained software.

那么整个比特币网络都依赖于核心客户端吗？作为比特币用户，你下载核心客户端后，其他软件或核心之外的东西会引用它吗？

And and so does does the entirety of Bitcoin work from in Core? Or if you're a Bitcoin user, you download Core, is it referenced in other software or things that exist outside of Core?

是的。所以我们，最近XE工具那个灾难性bug的问题就经常出现这种情况。没错，我们确实有一些依赖项，而且我们对这方面非常、非常、非常保守。其中一件事，就像迈克尔在构建系统中做的，他也非常注重减少依赖项数量。比如迈克尔长期以来发现的问题是，'哦，我们有个依赖项但没人维护它'。

Yeah. So we, this came up a lot with the recent like XE, utils, like, catastrophic bug. So yeah, we do have a few dependencies and we have we're like very, very, very conservative about it. And one of the things, again, like, like that Michael does as part of build system, he he's also very about like reducing the amount of dependencies. And so like what, for example, Michael has has found over over the time is that he's like, oh, we have a dependency and nobody's maintaining it.

或者基本上就是他们不再维护了。又或者它的代码审查标准达不到比特币核心代码应有的水平。不是所有代码都能被当作你自己写的代码来对待。

Or like, they they're just like yeah, they're not maintaining it basically. Or like it's not reviewed to the same standard that we would expect a Bitcoin core code to be reviewed to. And not all that code is like, you should treat it as the equivalent of your your code.

那个依赖项会放在哪里？

Where would that dependency sit?

会是GitHub上的另一个开源库。

It would it would be another open source library on GitHub.

嗯。那会不会形成中心化的力量？

Yeah. Would that be a centralizing force?

当我说它是开源的时候，中心化力量指的是如果他们删除了它

So, like, when I say it's an open like, centralizing force as in if they deleted it

之类的？是的。

or something? Yeah.

所以我们只是有一个它的副本，对吧？我们会使用自己的。而且我们必须

So we just have a copy of it. Right? And we would just use our own. And we have to

这么做。你想要所有东西都在核心中吗？那么有节点、外部依赖，还是那不可能？

do that. You want everything just in core? So there's node, external dependencies, or is that not possible?

抱歉。关于依赖方面，首先，我指的是构建比特币时引入的其他代码。不，当然

Sorry. I just I so dependencies wise, first of all, I was talking about, like, other code that gets pulled in when you build Bitcoin. No. Course

当它在运行时？

When when it's operating?

就像操作系统那样，你需要一个操作系统才能

It like, operating systems, like the like, you have you need an operating system in order

不，我是说，一旦你下载了核心，它要运行的话，是否需要实时引用其他系统，还是完全独立工作？

to No. I mean, like, once you've downloaded Core, does it for it to work, does it ever have to refer live to other systems, or does it work completely on its own?

哦，当你运行时会加载一个C库。但我想很难说...我们在这里讨论什么？你说的‘引用其他系统’是什么意思？

Oh, I you you load, like, a c library when when you run it. But I I guess it's hard to what are we talking about here? Like, is the what do you mean by refer to other systems?

比如，它是否依赖其他软件协议？还是所有功能都在核心模块内？

Like, are there other software protocols that it relies on? Or is everything within Core?

嗯，它需要...需要互联网。对，诸如此类的东西。

Well, it needs to it needs Internet. Yeah. Stuff like that.

我想说的是...

I'm on about like

对，抱歉。

yeah. Sorry.

可能我没表达清楚。我想问的是，你们是否需要维护另一个存放在服务器上的协议，比如AWS上的，需要引用某些功能时调用？

I'm probably explaining myself well. I'm on about, do you have to maintain like another protocol that's held on a server, AWS somewhere that has to refer to for things?

哦，我举个例子，DNS种子节点就是这种情况。当你启动节点时，怎么知道该连接谁？如何获取地址？我们不可能直接把地址...虽然源码里确实硬编码了一小部分，但这些地址通常每天都在变化。

Oh, I I'd say there is an example of that in the DNS seeds. So when you start up your node, how do you know who to connect to? How do you know addresses? We can't put that directly. Well, we have a little bit of that directly in source code, but usually that changes day to day.

对吧？所以你会连接一些DNS种子节点。大概有...我想说大概5个到10个之间，具体应该是8个，由不同的人运营。它们会告诉你：'我一直在扫描比特币网络收集地址，这里有一份非常多样化的地址列表'。

Right? So you will talk to, so some DNS seeders. There's, I wanna say, between five. I think there's eight of them, five to 10 of them, that are run by various people. And they'll be like, oh, yeah, I've been scraping the Bitcoin network for for addresses, and here's a list of addresses that are very diverse.

他们，你知道的，代表着世界上各种网络之类的。然后，这里有50个左右。你会和其中8个交流，然后就开始建立连接。这就是你正在连接的实时事物的一个例子。但我要说，相比你可能想到的大多数情况，这种情况非常少。

They're, you know, they represent various networks in the world or whatever. And then, you know, here here's 50 of them or whatever. And you'll talk to eight of these, and then, you know, you'll you'll start connecting. So that is an example of a live thing that you're connecting to. But I would say there's very little of that compared to most most things that you might think of.

好的。让我们来谈谈Iron，我们的下一代数据中心业务，利用100%可再生能源为比特币挖矿和AI的未来提供动力。非常喜欢这些家伙。Danny和我与创始人Dan和Will合作已经有一段时间了，现在超过一年了。我们对他们的价值观印象深刻，尤其是他们对当地社区和可持续计算能力的承诺，我们在各个方面都有合作。

Alright. Let's talk about Iron, who are our next generation data center business powering the future of Bitcoin mining and AI using 100% renewable energy. Love these guys. Now Danny and I have been working with the founders Dan and Will for quite some time, over a year now. We've been super impressed with their values, especially their commitment to local communities and sustainable computing power, and we've collaborated with them on everything.

我们制作了影片，录制了播客。他们甚至赞助了我的足球队Raoul Bedford，并且是我们会议cheat code的主要赞助商。Iron是为比特币挖矿、AI及其他领域提供动力的负责任方式。如果你想了解更多，请访问iren.com。

We've made films. We've done podcasts. They even sponsor my football team, Raoul Bedford, and they're the lead sponsor for our conference cheat code. Now Iron is the responsible way to power Bitcoin mining AI and beyond. And if you wanna find out more about them, please do head over to Iron, which is iren.com.

让我们谈谈比特币游戏领域的先驱BitCasino。他们成立于2013年，从那时起，BitCasino就成为游戏创新的灯塔。他们受到全球数万用户的信任，提供尖端的安全性，当然，他们支持世界上最好的货币——比特币。

Let's talk about BitCasino pioneers in Bitcoin gaming. Now they were established in 2013, and BitCasino, at that point, became the beacon of innovation in gaming and have been ever since. They are trusted by tens of thousands worldwide, and BitCasino offers cutting edge security. And, of course, they support the best money in the world. They support Bitcoin.

使用BitCasino，你可以享受闪电般的快速提款、VIP体验，以及超过4000款游戏。但他们也相信公平的游戏环境，提供零交易费用。他们提倡负责任赌博，并通过实时RTP功能让玩家识别看涨和看跌的游戏。如果你想了解更多并加入BitCasino，请访问bitcasino.io，那里顶级游戏与无与伦比的服务相遇。

Now with BitCasino, you get lightning fast withdrawals, VIP experience, and they have over 4,000 games. But they also believe in a fair gaming environment, offering no transaction fees. They promote responsible gambling, and they empower players with their live RTP feature to identify bullish and bearish games. Right. If you wanna find out more and join BitCasino, please head over to bitcasino.io, where top notch gaming meets unequaled service.

那就是bitcasino.io，也就是bitcasin0.io。同时，请记住要负责任地赌博。当你第一次拆分节点并尝试寻找其他节点交流时，是否有数量限制？比如只需要连接五六个就足够产生网络效应？还是可以连接所有节点？

That is bitcasino.io, which is bitcasin0.io. And also, please remember to gamble responsibly. When you split up your node for your first time and it's trying to find other nodes to talk to, is there a limit? It's like you only need to be connected to five or six, and once you've is that like the the that's enough to create the network effect? Or can it connect to all of them?

这类事情是如何运作的？好吧，我对你这些一无所知。

How does that kind of stuff work? Well I don't know any of this from you.

这里有几个设计目标，对吧？一个是带宽问题，对吧？首先你不可能连接到每个人，你的电脑很可能处理不过来。

So there's a couple design goals here. Right? One is like bandwidth. Right? You can't connect to every single person that just first of all, you'll probably like, your computer probably can't handle it.

而且这个确实，就是不可行。

And also this yeah. It's just not feasible.

因为他们需要维持持续的通信。

Because they're maintaining a constant communication.

哦，大概有几万个对吧？我确信你的电脑能建立的连接数量本质上是有限的。或者说抱歉...

Oh, there's like tens of thousands. Right? Yeah. I'm pretty sure you just like you have a limited number of slots essentially that you could connections that your computer could be making. Or sorry.

我试着用简单的方式解释，连接到比如说3万个节点是不可行的。但另一个设计目标是，你绝对需要首先找到诚实的节点，也就是那些一旦知道主链就会立即告诉你的节点，这样你才能找到工作量证明最多的链。我们理解工作量证明共识机制对吧？

I'm trying I'm trying to, like, make it easy to unders like, it's just not like, it's not gonna be feasible to connect to like, let's say 30,000 like nodes. Right? But the other design goal is like you definitely really, really want to find, first of all, honest peers, as in peers that are gonna tell you about the main chain as soon as they know about it so that you can find the most proof of work chain. Right? So like we understand proof of work consensus, right?

你会找到工作量证明最多的那条链，这就是你对链的定义。但你需要先得知它。所以节点通常会尝试维持8到10个出站连接，希望这些节点是多样化的、响应迅速的，并能及时告知主链信息。我们会收集所有连接节点的统计数据。

Like you find the chain that has the most proof of work on it. And that's your idea of what the chain is. But you have to hear about it. So yeah, what nodes will typically try to do is maintain, like eight to 10, yeah, let's just say eight to 10 like outbound peers that you connected to, and are hopefully diverse and hopefully very responsive and telling you about MainChain. So we'll collect statistics about everyone that we're connected to.

比如他们上次发送区块是什么时候？上次发送有效交易是什么时候？我发送ping后他们多快回复pong？所有这些指标都用来评估节点的质量。当我们管理已知节点的地址时，也会尽量追求多样性，对吧？

Be like, when's the last time they sent me a block? When's the last time they sent me a transaction that was valid? When's like, how fast do they Pong when I send them a ping? Like all of these metrics we use to assess like how good our peers are. And then when, you know, when we're managing all of the addresses of the nodes that we know about, we'll also try to go for diversity, right?

举个例子，如果有人想从某个ISP或其他地方快速生成一堆IP地址，这很容易。但要跨越多个不同的ISP这么做就困难得多。因此，我们才需要通过各种方式来确保我们连接到的是一组健康的节点。我们会定期监控，比如随机从地址簿中挑选一个节点，看看它是否会响应我们。

So it's very easy, for example, for someone to spin up a bunch of IP addresses from like one ISP or whatever. But it's much harder to do that across like a lot of different ISPs. So that's why, know, like we have all these ways of managing ways for us to make sure that we are connected to a healthy set of peers. And so, yeah, we'll like regularly monitor. Like, we'll regularly try like, okay, let's just like pick a random one from our address book and like see if that peer, is gonna respond to us.

如果某个节点连续几天都没有发送任何信息给我们——希望情况不至于那么糟——那基本上就是个糟糕的节点。我们会将其剔除并尝试连接新节点。你看，软件中涉及的远不止工作量证明链或签名验证这些内容，整个系统非常复杂且迷人，真的非常有意思。

And and like if if we have like one of our peers has not sent us anything for, you know, days, hopefully not not at that point, like, you know, is is a shitty peer. Like we'll evict them and we'll and we'll try a new one. And so, yeah, there's there's so much that goes into the software that is not just, you know, what's the most proof of work chain or like is the signature valid? And it's very complex and it's very interesting. Very, very interesting, I think.

我在大学上网络课程时完全不了解这些，现在想想确实非常...

Like, I didn't I didn't know about this when I was taking a a networking class in in college, in uni. And, yeah, it's very

很酷。那么它们之间主要共享什么信息？是让网络运作的交易数据和区块吗？

it's cool. And and what and what are they sharing with each other? Is it is the primary information being shared to make the network work transactions and blocks?

我认为主要是交易数据、区块以及节点地址。

Well well, I would say it's like primarily transactions, blocks, and addresses of nodes.

好的。

Okay.

嗯，还有...

Well, there's

因为我觉得有很多人，大多数人可能根本不知道比特币是如何运作的。是的。因为很多人甚至不会下载一个节点。只有极少数人会运行节点。

because I think there's a of people this most people probably have no idea how Bitcoin works. Yeah. Because a lot of people don't even download a node. It's very small percentage of people will operate a node.

每个人都应该运行一个节点。你会学到很多关于比特币的知识，而且非常有趣，也非常简单。运行一个节点真的非常非常简单。

Everybody run a node. You'll learn so much about Bitcoin, and it's so fun, and it's so easy. It is so so easy to run a

我们会让你大开眼界。

We make a brain blow you.

是的。你不需要，甚至不需要600GB的空间。这是个误区。人们普遍认为你需要600GB来存储区块链，这是个非常常见的误解。你只需要存储JIXFit部分。

Yeah. We you don't need like, you don't even need, like, 600 gigabytes. Everybody like this is a myth. It's like a very common myth that people think you need 600 gigabytes to to store the blockchain. You only need to store the JIXFit.

对吧？你不需要那么多空间。你可以在笔记本电脑上运行它。

Right? Like, you don't need that much. You can run it on your laptop.

但最主要的是地址、交易和区块。是的。所以当一笔交易被创建时，它会在整个网络中传播。

But the primary thing is addresses, transactions, and blocks. Yeah. And so when a transaction is created, that is relayed across the whole network.

是的。它永远不会

Yes. It's gonna never be

有时候我的节点就是无法接收到那笔交易。还是说最终它总会收到？

a time when when my node just doesn't pick up that transaction. Or will it eventually get that transaction?

所以我们讨论的是尚未被挖矿确认的交易

So we're talking about transactions that haven't been mined in

在一个区块里

a block

对吧？没错。这其实是我最喜欢讨论的话题之一。

yet. Correct? Yeah. So, yeah, this is like what I love talking about the most.

内存池。

Mempool.

是的。问题在于：你是否一定能得知交易信息？绝对可以。实际上我参与的许多项目都在不同层面上涉及交易审查规避。我刚描述的是点对点的例子，但还有内存池层面的例子。关键是要避免设计出容易被狙击或干扰的系统。

Yeah. So the question was, are you guaranteed to hear about transaction? Absolutely And some some things that I basically, the projects that I work on, a lot of them kind of censor around transaction censorship on various different levels. So I kind of described a peer to peer example, but there's also mempool examples. And there's also like, don't design your things in a way that, you know, it can get sniped or whatever.

这种情况在Mumples也可能发生。但我们的目标是建立一个抗审查的支付网络，任何人都能自由交易且无法被阻断。我认为这是比特币三大最佳特性之一，也是我们应努力维护的。所以没错，虽然不能百分百保证，但或许我该聊聊内存池相关的内容？

There it can happen in Mumples as well. But the the goal is like we're trying to build a censorship resistant payment network where anyone can pay each other and nobody can block that. I think this is like top three best things about Bitcoin or things that we should strive for in Bitcoin. And and so, yeah, you aren't guaranteed. So should I talk about, like, Mempool stuff?

是的。

Yeah.

完全没有？好吧。是的。

At all? Okay. Yeah.

所以，你看，如果我没接收到交易，那它就不那么重要。但显然，每个构建区块的矿工都能看到每笔交易。但矿工有可能看不到某笔交易吗？或者说，大家的交易池应该相同吗？每个人都应该有相同的交易池吗？

And and so, look, if if if I don't pick it up, it's not so important. But it's important, obviously, every miner who's constructing blocks sees every transaction. But it are there chances that a miner might not see a transaction? Is the or should the member be the same? Should everybody have the same member?

感谢你提出这个问题。这确实是很多人讨论的话题。显然，比特币是可以正常运行的。之前有很多讨论旨在告诉大家，并非所有人的交易池都相同。不存在所谓的'唯一交易池'。

Thank you for for asking that question. This has been a topic that a lot of people talk about. Obviously, Bitcoin could function if okay. So there was a lot that was said to kind of teach everyone that not everybody has might have the same mempool. There's no such thing as the mempool.

对吧？网络中的每个节点都有自己的交易池，它们可能相同也可能不同。我认为当系统运行良好时，希望大家的交易池是一致的。我会解释原因：因为我们有这个去中心化的交易转发网络。

Right? There's each node on the network has their own mempool and they may or may not be the same. I think when things are working really well, hopefully everyone like does. And I'll explain why. So like, we have this decentralized transaction relay network.

每个节点都有一个交易池。

Each node has a mempool.

我们需要先解释一下什么是交易池吗？

Should we should we explain what the mempool is?

好的。当然。

Okay. Sure.

是啊。因为新听众刚进来时可能会想，你们他妈的在说什么？他们能听到这里已经很不错了。他们会疑惑，你们俩在聊什么？Gloria到底在说什么？

Yeah. Because the the we've got new listeners coming in just they're like, what the fuck you wanna they're probably actually, if they've got this far, well done. They're gonna think, what are you two talking about? What's Gloria talking about?

好的。对。抱歉。如果没问题的话，我们从头开始讲吧。

Okay. Yeah. Sorry. If okay. So let's start from the beginning.

我们需要抗审查的支付系统。实现这一目标的方式是通过网络的去中心化结构。明白吗？我们不是简单地让矿工运营一个网站来接收付款，那样就毫无意义了。

We want censorship resistant payments. The way that we achieve that is through the decentralized structure of the network. Right? We don't just have, you know, the miner has like a website that you submit payments to. That would not be sense.

或者政府或银行运营的网站，对吧？显然——希望大家都明白——那行不通。所以我们采用的是去中心化网络。你连接到数万或数十万个节点中的一个，向对等节点发送交易，它们会进行传播，最终到达所有节点。这才是应有的运作方式。

Or the government has a website where or a bank, right? That's that's obviously that doesn't hopefully everybody understands that that's not how that so what we have is decentralized network instead. You connect to one of the tens of thousands or hundreds of thousands of nodes. You send your peers your transaction and they gossip it and eventually it gets to everybody. That's how it's supposed to work.

实际实现方式是：每个节点首先会接收并验证交易，将其存入内存池（mempool）。这只是未确认交易的缓存区，就是个缓存而已。

How we implement that in practice is each node, first of all, is down to hear about transactions and validate them, and we'll put them into their memory pool or mempool. And that's just a cache of unconfirmed transactions. It's just a cache.

你能用垃圾信息塞满那个缓存吗？然后...

Can you spam that cache and therefore, there you

好问题。嗯，是的。作为一个节点操作员，你为什么要这么做？一个原因是为了在区块到来时能缓存所有内容。

Good good question. And okay. Yeah. So you as a node operator why do you why do you do this? One is so that you can cache everything for when blocks come.

比如其中一个原因是为了让区块中继工作。当你收到一个区块时，如果之前没看到任何交易，可能会发生的情况是网络带宽激增，你需要下载整个区块。然后你会不停地运转、加载UTXO（未花费交易输出）从磁盘，进行验证所有签名等计算密集型操作，这非常耗费计算资源，可能需要很长时间。

Like one of the reasons is so that Block Relay works. So when you receive a block, what you what might happen if you didn't see any transaction before is you have this huge network, like bandwidth spike where you're downloading the full block. And then you're like spinning, spinning, spinning. You're like loading UTXOs, from disk and you're like like doing the computationally heavy thing of verifying all the signatures, etcetera. It's very computationally expensive and like it might take you a while to do this.

所以这是一种可能的情况。另一种可能是：是的，我已经下载并验证了所有这些交易，我有一个缓存存储了所有检查过的签名，已经把UTXO都加载到内存中完成了，然后存储到下一个区块。

So that's one possible reality. And the other possible reality is like, yep, I already downloaded all these transactions, already verified them. I have a cache where I like cached all the signatures that I checked. I already loaded all the UTXOs into memory and I'm done. And I store it onto the next block.

对吧？这对你作为节点操作员来说首先是个更健康的状态，对网络也是，因为每一跳都快得多。抱歉，区块延迟就这样大幅降低，因为每一跳都会成倍影响。明白吗？

Right? And that is a much, much healthier, first of all, situation for you as the node operator, but also for the network because every single hop is so much faster. Sorry. So block latency goes from like this to this because each hop, like multiplies at each hop. Right?

所以最健康的网络是重组最少的，因为重组很糟糕。比如你希望区块延迟非常快，因为如果两个人几乎同时发现下一个区块，最好大家都能快速获知其中一个。否则可能会出现一半网络知道这个区块，另一半知道那个区块，最终不得不解决这个重组。重组越少越好。

And so the healthiest network, the ones with the most the least reorgs where like reorgs are bad because So for example, you want block latency to be very, very fast because if two people find the next block at around the same time, hopefully everybody hears about one like quickly. Otherwise, you might end up with, you know, half the network knows about this block and the other half knows about this block. And then at some point, you have to resolve that that reorg. The fewer of those you have, the better.

对吧？这种情况多久发生一次？

Right? How often do they happen?

我想大概每周一次吧。

I think like once a week or something.

哦，所以它还是挺规律的。

Oh, so it's still quite regular.

我不确定是不是那样。大概每千个区块左右吧。嗯。对。不过以前更频繁些。

I don't know if that yeah. Every thousand blocks or something. Yeah. Yeah. But it used to be more frequent.

每次发生这种情况，比如每次有重组时，就会很混乱，意味着你确认过的支付可能不再被确认。这是个用户界面问题，影响使用体验。对吧？

And every time that happens, like every time there's a reorg, like it's it's messy, it means your confirmed payment may or may not still be confirmed. Like it's a UI issue. It's like a it's a usability issue. Yeah. Right?

是啊。

Yeah.

所以这是个安全问题，至少可能是。因此我们真的想尽可能减少这种情况。总之，我们刚才在讨论为什么内存池和交易转发对区块转发有用。另一个作用显然是为了建立一个抗审查的支付系统。

So it's a security issue. It can be at least. And so we really we really want to minimize that as much as possible. So anyway, we were talking about why Mempools and transaction relay are useful for block relay. And the other use is obviously so that we can have a nice censorship resistant payment system.

当我们参与交易转发时，不只是验证交易是否符合共识规则就扔进内存池继续。你刚才问的关于能否向安全内存池发送垃圾交易的问题——确实存在需要几分钟来验证的共识有效交易，对吧？

And when we are participating in transaction relay, this is we don't just like validate whether the transaction is consensus valid, chuck it in mempool and and continue. So your question which you asked, which was like, can you spam the secure mempool? Yeah. There are consensus valid transactions that can take you several minutes to validate. Right?

这些交易可能是有效的也可能不是。记得Rusty Russell很久前写过一篇很棒的文章，关于一个非隔离见证交易因二次缓存问题导致需要几分钟验证的例子。但你可以想象那可能是伪造交易。

And they may or may not be a valid transaction. Right? Like there's this really great post, think by Rusty Russell, which is from a long time ago, a non SegWit transaction with like a bunch of inputs due to like quadratics caching. It took like several minutes to validate. But you can imagine that being like a fake transaction.

对吧？那是在一个包含工作量证明的区块里出现的。工作量证明这种机制能轻松防止垃圾信息，因为生成一个有效的工作量证明区块比下载验证它要困难得多。而交易则不需要工作量证明，所以更像是有人随便发个交易给你，你可能花几分钟验证后才发现它是无效的。

Right? That came in a block, which had proof of work on it. And proof of work is this like really easy way to avoid spam because it's so much harder to produce a proof of work valid block than it is to like download it and validate it. Whereas transactions, you don't have a proof of work requirement. So it's much more like people who just send you this transaction, maybe you spend minutes validating it and then it turns out to be invalid.

这完全没用。甚至都不会被确认。对吧？嗯，谢谢。

It's completely useless. It's not even gonna be confirmed. Right? Yeah. Thanks.

谢谢。对吧？所以我们制定了规则，比如不接受超大额交易——这就是个很简单的共识规则。理论上可以有40万权重单位的交易。

Thanks. Right? And so we have rules where we're like, okay, we don't take really large transactions. Like that's a very simple rule. So consensus rule, you could have like a 400, 400,000 like weight unit like transaction.

但我们在交易转发时限制为10万。另一个典型例子是内存池——它就是个缓存区，对吧？我们希望通过大量激励相容指标来评估交易是否有用。

But we make that 100,000, in in transaction relay. Another really good example is, we use like we want our mempool, it's a cache. Right? We want it to be useful. And so we use a lot of incentive compatibility metrics to assess whether or not this transaction is useful.

最典型的例子是遇到两个冲突交易时，它们花费同一个UTXO形成双花。这时应该选择手续费率更高的那个，因为对矿工来说激励相容性更强。即使你不是矿工，在验证区块时也会倾向于选择更可能被打包的交易。

So the best example is if you get two transactions that conflict with each other, they spend the same UTXO, they're double spends. You you should probably pick the one that's higher fee rate. Because that is more incentive compatible for a miner to mine. Even if you're not a miner, you like, you're expecting to use this for when you're validating blocks. You're like, okay, this is more likely to be mined and therefore I'm gonna take it.

这就是我们在交易验证中的另一个实践，也是实现费用替代（RBF）的基础。通过这种方式提升手续费。抱歉。

So that's like, that's another example of like what we do in transaction validation. And that's what leads to being able to RBF, to replace by fee. Right? To fee bump that way. Sorry.

我正要回答你的问题。关于是否每个人都该...希望如此

I'm getting to your question Yeah. About should everybody have the same Like, hopefully

在一个理想的世界里。

In an ideal world.

是啊。在理想情况下，如果每个人都有相同的内存池，事情会运作得最好。当然，像我描述的所有这些规则，它们并非共识。所以你可以随心所欲，对吧？

Yeah. In an ideal world, things work best if everybody has the same mempool. Of course, like all these rules that I described, they're not consensus. So you can have whatever you want. Right?

比如假设你说，我的内存池——我是说，举个例子，我讨厌铭文。或者我的内存池不允许花费已确认的UTXO，或者我的内存池...你想怎么做都行。是的，那样是被允许的。但这是否意味着因此每个人都该有不同的内存池？其实不是。

Like let's say you're like, my mem like, I mean, the example is like, I hate inscriptions. Or like my mempool like doesn't allow spending on confirmed UTXOs or my mem Like you can do whatever you want. And that, yes, like that's allowed. But is that But like people are like, therefore everybody should have different mempools. It's like, well, no.

那不是我们的初衷。但我觉得这是对过去观念的一种过度修正——人们曾以为存在一个统一的内存池来提交交易。而现在大家突然意识到，实际上每个人都有自己的内存池，所有内存池都该不同。

That's that's not where we were going. But I think it was like kind of this overcorrection of like people used to think there was the mempool that you would submit transactions to. And then now everybody is like, oh, actually, everybody has their own mempool. All the mempool should be different.

我以前也这么认为。真的。曾以为只有一个内存池。所以我才会问是否存在外部依赖——因为我原以为存在一个统一的交易内存池。

I used to think that. I did. Used to think there was just like one mempool. That's why I asked the question, are there outside dependencies? Because I imagine there was this this mempool of transactions.

我原先没意识到交易本身是被转发和存储的。所以当我们访问Mempool网站时

I didn't realize the transactions itself were getting relayed and stored. So when we go to the Mempool website

嗯。Mempool是

Mhmm. Mempool Is

那只是他们自己的

that just their own

内存？是的。是的。是的。理论上来说，内存池空间和内存池是不同的东西。

memory? Yeah. Yeah. Yeah. That's Theoretically So Mempool space and Mempool are different things.

每个节点都有自己的内存池MEMPL。是的。然后内存池空间就形成了。

There's like MEMPL the memory pool that each node has. Yeah. Then MEMPL space happens.

是的。所以那只是理论上的。那只是它应该的样子。

Yeah. And so that's just theoretical. That's just what what it should be.

我认为不一定。

I think not necessarily.

他们是从哪里获取的？

Where did they take theirs from?

所以他们为什么不...我是说，他们运行自己的比特币节点。是的。

So they why don't so I mean, they run their own Bitcoin nodes. Yeah.

这是我的假设。但因此，理论上他们的可能与我的不同。

That's what my assumption is. But but therefore, theirs could theoretically be different from mine.

是的，是的，完全正确。首先，我认为他们运行了几个不同的节点。他们有一个包含所有交易的大型节点，还有一个默认节点，容量是300兆字节。

Yes. Yes. Absolutely. So I I I think they run a few different nodes, first of all. And so they have like a huge node that has like all the transactions and then they have like the default node, right, which is 300 megabytes.

所以当费率很高、交易量巨大时，大型内存池会保留所有交易。但比如你的300兆内存池会开始清除那些费率较低、不太可能被打包的交易。我们正利用激励相容性来评估...

And so like when fee rates are really there's like a ton of transactions, like the huge mempool will have all of them. But for example, your 300 megabyte mempool is gonna start purging the lower fee rate stuff that is less likely to get mined. We're using incentive compatibility to assess what

那当它们被清除后，会去哪里呢？

So when they get purged, where do they go to?

直接丢弃。它们就消失了。从我的节点？对你的节点来说。我们直接丢弃它们。

You just throw them away. They're gone. From my node? For your node. We'll just throw them away.

但矿工永远不会丢弃任何交易，因为说不定哪天就会打包它们？

But a miner would never throw any away because at some point they might mine them?

我猜他们的做法和Mempool Space很相似——用大型节点记录所有交易，但保留一个默认节点。因为内存池越大，所有操作的计算成本就越高。比如构建新区块模板时，你更希望基于较小的内存池操作。

I I would imagine that they do something very similar to what Mempool Space does, which is they have a huge node that just keeps track of all of them, but they have like a default node. Because the larger your mempool, the more computationally expensive everything gets. Like building a new block template, for example, you wanna do that on a smaller mempool.

所以那个巨大的存储空间就是用来存放所有后续的东西。就像，如果我们经历一个交易量很大但价值很低的密集时期，他们就把这些交易暂时存放起来，等手续费率下降后再处理？

So the huge one is just for storing all the stuff later. It's like, okay, if we go through a very intensive period where there's lots of low value transactions, they park them for later and then, you know, fee rates drop, then they bring those in?

对，正是这样。

Yeah. Exactly.

今天倒没那么多。

Not so much today.

没错没错。而且我觉得Mempool Space那个审计功能挺酷的，他们会自己构建区块模板，然后和实际挖出的区块对比，这样就能看出差异——要知道这不一定是审查，交易不同的原因可能有很多种。

Yeah. Yeah. Exactly. And and so I I think part of like, Mempool Space has this cool audit feature, right, where they like they build their own block template and they compare it to what got mined and you can see like, oh, are these, you know, it's not always censorship. Like, there's there's many reasons why the transactions might be different.

但你能大致看出原本应该是什么样子，或者差异在哪里。不过当然，你的节点可能和MemplSpace的节点不同，而矿工的节点可能又不一样，所以很难确切说明交易为何不同。

But like you can kind of see what the what it would be or like what the difference was. But of course, like your node might be different from MemplSpace's node and, you know, and that might be different from the minor node or whatever. And so it's like very difficult to it's very difficult to prescribe a reason why a transaction might be different.

有没有可能我发起的交易手续费太低...嗯...结果永远不被打包？就这么一直挂着直到天荒地老？

Could could I ever post a transaction where the fee's too low Mhmm. And it never get mined? And it just sits there and lasts forever.

这个嘛，如果所有人出的区块空间竞价都比你高，那抱歉你就只能...我能...

Well, if, like, everybody's bidding higher than you for block space, then sorry you got you got Can I does

那笔交易最终会被取消吗？我可以哦？需要我自己想办法取消吗？我要把它退回去吗？具体是怎么操作的？

that transaction eventually just get canceled? Can I Oh? Do have do to I have to somehow cancel it myself? Do I bring it back? Like, how does that work?

我认为你无法取消它。一旦你签署了交易，比如矿工可能会记录他们看到的每一笔交易。六个月后，你以为交易取消了，可能你正在给朋友或商家付款，然后你说，好吧，我再发一笔新交易给你。

So I don't think you can cancel it. I think once you've signed that transaction so for example, a miner could just be logging every single transaction that they see. And then like six months later, you thought it was canceled. And maybe you were like paying your friend or a merchant or something. And you're like, okay, I'll just send you a new transaction.

那笔可能被挖出来。所以很抱歉。

That one's canceled. Like, you might get mined. So sorry.

你怎么把它退回去？有没有办法能退回去？

How you how do you bring it back? Is there any way you can bring it back?

不，不行。抱歉，没办法。不过你可以尝试双花。

No. No. Sorry. There isn't. Well, you could double spend it.

对吧？好的。但这仍不能保证别人不会挖到它。对吧？这是一笔共识有效的交易。

Right? Okay. It's still not guaranteed that someone might not mine it. Right? It's a consensus valid transaction.

你已经广播出去了。

You've already broadcast it.

我可以提高手续费吗？

Can I increase the fee?

可以，可以，可以。就像这样。你可以像我之前描述的那样设置手续费，你可以创建一个RBF（费用替换）交易，用更高费率的交易进行双花。

Yeah. Yeah. Yeah. Like so yeah. So you could like fee like I described before, you could create an RBF where you double spend it in a higher fee rate transaction.

理论上，矿工应该会选择费率更高的那个。但同样，如果他们愿意，他们可以决定——如果他们表现得很奇怪，比如他们不知怎么没听说手续费提高了，或者在你提高手续费时，他们已经开始打包包含原交易的区块了。这些都是可能导致你的RBF交易成功或失败的原因，有很多因素会影响费用替换交易能否生效。

And theoretically, a miner should mind the higher fee rate one. But again, they could decide if they wanted to, if they're being weird, like they didn't hear about, like, you know, for some reason they didn't hear about the fee bump. Or by the time you fee bumped it, they had already started working on a block with that transaction in it, for example. These are all potential reasons why like your RBF may or may not get in, may or may not happen. There's a lot of reasons why the replace by fee transaction may or may not, work.

其中一些原因——因为我花了很多时间研究RBF的局限性——是我们的策略规则不够智能。我们再次在抗拒绝服务攻击（比如保持计算复杂度低）和激励相容性之间做出权衡。有时我们会走捷径，用不完美的启发式方法来评估激励相容性，因为精确计算的计算量太大。所以我们可能会看着两笔交易说‘哦，这笔更适合挖’，但实际上另一笔才是。

One of those like some of those reasons, because I've spent a lot of time looking at kind of the limitations of RBF, is just like our policy rules are not very smart. So we make a trade off again between comp like being resistant to denial of service, like trying to be computationally not complex, as well as incentive compatibility. Right? So, sometimes we make a trade off where we take a shortcut or we have an imperfect heuristic for assessing incentive compatibility because the actual thing would be too computationally complex to do. And so we might look at two transactions and we'll be like, oh, this one's better to mine, but actually this one is.

因此我们可能会拒绝实际上更好的替换交易，或者接受不够好的替换。我认为这是个漏洞，而且非常复杂，因为这是大量工程决策的结果，属于极端案例。确实，我花了很多时间研究这类RBF策略的局限性。我和其他人也提出了一些改进方案——当然我只代表我个人观点。

And so we might reject a replacement that's actually better or we might accept a replacement that's not better. And like this this is I consider a bug. It's a very complicated bug because it's the result of kind of a lot of engineering and the and these are very like kind of edgy edge cases. But yeah, there are I've spent a lot of time essentially working on these kinds of replace by fee policy limitations. And we have a few I have some proposals as well as other people have I don't speak on behalf of anyone other than myself.

目前有很多令人兴奋的项目正在努力修复内存池（Mempool）中的这些问题。

There there are a lot of really exciting projects that people are working on to try to fix these things in Mempool.

那么铭文（Inscriptions）是如何改变内存池规则的？显然它已经造成了影响。

So how has Inscriptions changed the Mempool game? Because clearly, it has.

我记得我们有过三轮辩论。大概就是人们因为BRC20相关的事情对高昂的手续费非常不满。

So I remember we had like a three round trip debate Yeah. Where I think it was around, like, people were really upset that the fees were so high because of all the, I think, BRC 20 stuff.

是的。

Yes.

他们当时问能不能解决这个问题。我就说，这就像你在高速公路上遇到堵车，却质问施工人员为什么让这么多车通行。但实际上，其他人当然有权发送比特币交易。交易能否被确认取决于手续费率，因为矿工有经济动力优先处理高手续费交易。他们可以利用区块空间...

And they were like, can you fix this? And I was like, you're kind of like somebody who's driving down the highway and there's traffic and you're like, road workers, why did you allow all these people? And like, it's like, yeah, other people are allowed to are allowed to send Bitcoin transactions. And the way that they're chosen for confirmation is based on fee rate because miners are, you know, economically incentivized to include those transactions. And they can use the blocks yeah.

没错。他们是在利用稀缺的区块空间应对高需求，这是经济规律。区块空间这种资源的价格会随着需求上涨，这很正常。

Yeah. I mean, they use the scarce block space to a, you know, a demand that's very high and it's using economic. Right? Like, the price goes up for a unit of block space, which is the resource that you're trying to consume.

但这些问题带有政治色彩。

But this stuff is political.

确实，这是...

Yeah. It's it's

作为开发者或维护者，你是否觉得自己有责任保持中立？只要保证软件正常运行，就不该表达个人立场？

And and do you feel like, as a developer or a maintainer, either or, do you feel like you have a duty to not be political? Not have an opinion as long as the your job is just to make sure the software works.

是的。我们讨论这个问题时，我遇到的困扰是人们在问：你能阻止垃圾信息吗？而垃圾信息在我们谈论内存池代码和交易转发代码时有非常明确的定义——就像我们刚才讨论的那样，它指的是不当占用资源（比如你电脑的计算力或内存）的行为。我们有一整套防范机制。因此，在处理这类代码时，我采用了自己对垃圾信息的定义标准。

Yeah. So the issue that I had when we were talking about this was that people are like, can you stop the spam? And spam has a very specific like we've just been talking about the definition of spam when we're talking about mempool code and transaction relay code, where, you know, it's it's using a resource, for example, your computer's, you know, computation or memory, that it shouldn't be. And and we have all kinds of ways of preventing that. And so we have a I have a definition of spam that I use when I'm work when I'm thinking about this kind of code.

但人们要求我根据具体应用场景来适用这个定义。我认为协议开发者或交易转发代码的职责并不包括判定哪些应用场景是合理的。你懂我的意思吗？

And people are asking me to apply that definition based on use case. And I don't think it's the protocol developer, the transaction relay codes place to decide what use case is legitimate or not. Like it like, you know what I'm saying?

这个决策可以由节点甚至矿工层面做出，就像Ocean所做的那样。在去中心化网络中，你确实可以自主决定。

It can happen at a node or even minor level like Ocean have done. You can that's where you can make that decision because in the decentralized network Yeah. You get to decide yourself.

没错。就像你不喜欢猫咪照片或扭扭糖（Twizzlers）是你的自由。但我不认为应该用'这不是合法交易'或'NFT糟蹋区块空间'这类说辞——这种价值判断不该出现在编写策略代码的考量中。

Yeah. Like you like if you don't like cat photos, you don't like twizzers or whatever, that's like your choice. Right? But I don't think this is not a legitimate transaction or, you know, this is a waste of blocks, you know, because NFTs are bad or whatever. I don't think that language has a play Well, I don't think that should be considered when you are talking about writing policy code.

确实，作为开发者...

Yeah. As a as a developer.

对，作为开发者，作为协议维护者...

Yeah. As a developer. As a protocol maintainer.

这个话题确实很难把握。说实话我反复思考过，现在更倾向于比特币应该只用于金融交易——毕竟比特币是货币。虽然也有人辩称这些（非金融）应用本质上还是...

Like It's been such a tricky subject to navigate. And honestly, I've gone back and forth on this. I've come to the point that my preference would be Bitcoin was just used for financial transactions Sure. Of Bitcoin being money. And and I know someone said, well, these still are.

但我也同样不喜欢审查制度的概念。我和比特币机械师进行了长时间交谈，显然他深度参与了海洋项目。我赞同他的观点——我真正关心的是货币体系的改进，特别是在马拉维、肯尼亚和加纳等地目睹的那些情况。

But I've also I don't like the idea of censorship. But had a very long conversation with Bitcoin Mechanic. Obviously, he's very much involved in Ocean. I agree with him. The like, the thing I care about, the thing I'm here for is the improvement of money, especially with the kind of things we saw out in Malawi and Kenya and Ghana.

是的，我对量子猫之类的东西毫无兴趣。我觉得其中某些内容是对比特币的攻击。但我逐渐意识到这让你陷入两难境地，因为人们希望你表达政治立场。你是否感受到这种压力？

Yeah. I have zero interest in in in quantum cats. I feel like some of that some of this stuff's an attack on Bitcoin. But I I've I've come to really appreciate that it puts you in a difficult position because people want you to have a political view. It just is that do you feel that as a pressure?

确实如此。不过与其说人们要求我表态，不如说他们是在要求我们修改软件规则。公平地说，我们已经找出了几个需要在内存池设置非共识性验证规则的政策性理由。

Yes. Yes. Well, I don't know if people are asking me to have a view so much as they're demanding that we change the software and and right? Yeah. And I I think let's to be kind of fair to this argument, we've identified a couple of reasons why you would have policy, policy as in the validation rules in Mempool that aren't consensus.

历史上确实存在一些限制措施来防止消耗网络资源的行为。比如我们设有标准输出类型和粉尘限额，这些机制旨在保护网络——包括见证折扣政策也是为了防止UTXO集合过度膨胀。

And historically, there have been, kind of things put in place to discourage things that are, bad for network resources. So for example, we have standard output types and we have a dust limit. And that's to protect try to protect or try to discourage, creating a lot of and even the witness discount as well is kind of to try to discourage bloating of the UTXO set.

什么是见证折扣？

What is a witness discount?

哦等等，我们先把这个话题讲完。我刚才说到运行完整比特币节点其实不需要存储整个区块链，

Oh, no. We're getting into it. Okay. So let me just finish that idea first. So I talked about how you can run a fully validating Bitcoin node and you don't need to store the entire blockchain.

只需要存储UTXO集合就够了——当然还需要区块头链。你说得对，确实不需要存储全部数据。

You just need to store the UTXO set, all of the UTXOs. And that is enough for you to validate whether each transaction and each block is valid. Well, you also need the headers chain. Yeah. You don't need all you're right.

你可以修剪大部分内容。因此，UTXO集的大小对于扩展性、运行节点的可访问性等方面非常重要。你刚才说什么来着？哦，见证折扣。当我们查看交易大小时，交易的某些部分权重与其他部分不同。

You can prune most of it. And so the size of the UTXO set is kind of very important to scaling and to the accessibility of running a node and all those things. And so what were you saying? Oh, witness discount. So when we are, looking at the size of a transaction, certain parts of the transaction are are weighted differently from others.

见证数据（如签名或使交易有效的脚本）可以被丢弃，这部分权重较低。这就是见证折扣的全部含义。这意味着如果你创建大量见证数据，费用不会像创建大量输出（这些输出会进入UTXO集）那样高昂。同样，我们有一系列标准输出类型，比如支付给见证公钥哈希、支付给Taproot或支付给见证脚本哈希等标准类型。

And so the witness data, which is like the signatures or the scripts that make this transaction valid, which can be discarded, like that is weighted less. That's all the witness discount is. And so that means that if you make a large witness, that does not cost that's not as that won't cost you as much in fees, for example, as if you like made a lot of outputs and outputs are what go into the UTXO set. And similarly, like we have a list of like standard output types. So that's paid to witness pubkey hash or paid to taproot or paid to witness script hash or, you know, those like standard types.

它们非常小且具有标准格式。至于粉尘限制我们就不深入讨论了，但存在各种...

And, you know, they're they're very small and and they have a standard format. And say, okay, we don't have to get into the dust limit. But these are there are like various

粉尘限制到底是什么？实际数值是多少？

What is the dust limit though? What's the actual value?

当然。这取决于输出类型。我刚描述了不同的输出类型，不确定是否需要解释具体含义。但如果你创建一个仅含1聪的输出，当你花费这个UTXO时，手续费会超过其实际价值。

Sure. Oh, it depends on the output type. So I just described the different alpha types. So I don't know if I should explain what they are. But if you create an output that has like one Satoshi in it, well, you're not like, if you are to spend that, it's going to cost you more in fees to spend that UTXO than it's worth.

你明白我的

You know what I'm

意思吗？除非你持有特殊聪币。比如第一枚聪

saying? Unless you've got a special Satoshi. The first Satoshi

哦，当然。当然。

Oh, sure. Sure.

因为这完全是胡扯，对吧？你无法识别单个聪。

Because that that's that's bullshit. Right? You can't identify individual Satoshis.

嗯，这就是Ordinals的原理，对吧？就像你有一个外部的，像是

It well, that's what Ordinals is. Right? Like, you have an external, like

但他们基本上只是编造了自己的编号系统，这些聪并不独立存在。因为UTXO本质上就像——如果我把一张20英镑的钞票放在桌上——嗯，里面并没有单独的便士。它就是一张20英镑的钞票。你无法识别其中单独的便士。确实。

But they've they're like they've just basically made up their own numbering system, but these Satoshis don't exist individually. Because a UTXO is is essentially a a like, if I put a £20 note on the table Mhmm. There aren't individual pennies in there. It's just a £20 note. You can't identify the individual pennies in the Sure.

我猜我们可以某种程度上

I I guess we can kind of

不，你不需要那么做。

No. You don't need to do that.

我...我觉得是的。就像，你可以对协议本身未内置的含义做任何解读。但Ordinal...

I I think yeah. Like, you can like do anything to to derive meaning that isn't kind of baked into the the protocol. But ordinal

理论是指比特币之外的外部协议。

theory is something that says an ex an external protocol outside of Bitcoin.

是的。这没什么好说的，当人们叫我停止搞序数时，我们根本没有

Yeah. That has nothing to like, when people tell me to stop doing ordinals, we don't have

这跟序数没关系。去你的吧。没错。所以那个粉尘限额嗯哼。

It's got nothing do. Ordinals in. Fuck yourself. Yeah. So so the dust limit Mhmm.

这个限额是否需要随着比特币价值变化而调整？还是说因为它是以聪计价就无关紧要？

Does that have to be a kind of a moving beast as Bitcoin value changes? Does or is it irrelevant because it's priced in Satoshis?

它是按聪为单位确定的。对吧？假设你总是需要每虚拟字节至少支付1聪。对。小数点后已经是三位了。

It it's priced it it's determined based on like Satoshi units. Right? So if you'd if you assume that you let's say, you're always gonna have to have at least one SAT per V byte. Yep. The decimal is already three.

但如果每笔交易都要支付1聪/虚拟字节才能花费，那你就不能创建比如5聪的输出，因为这种输出的价值还抵不上使用它的成本。

But like one SAT per V byte on any transaction in order to spend it, then you can't be making like if you make a five SAT output, then like you prob like it's it's uneconomical to create that output because it's worth less than what it would cost to use it.

好的。我们来聊聊Ledger，我最爱的硬件钱包。从2017年用到现在。Ledger是比特币安全领域的全球领导者，也是你掌控私钥的最佳方式之一。听着。

Alright. Let's talk about Ledger, my favorite hardware wallet in the world. The people I've been using since 2017. Now Ledger is a world leader in Bitcoin security, and it's one of the best ways for you to own and secure your private keys. Now listen.

如果你还在交易所或托管机构持有比特币，你在干什么？我多年来一直告诫你不要这样做。是时候更认真地对待你的资产安全了。记住，若非你的私钥，比特币就不属于你。现在通过Ledger硬件钱包搭配Ledger Live应用，你拥有了最便捷的方式来管理自己的私钥并掌控比特币。

If you are still holding your Bitcoin on an exchange or with a custodian, what are you doing? I've been telling you for years to stop doing this. It is time to take your security more seriously. Remember, if it's not your keys, it ain't your Bitcoin. Now with Ledger hardware wallets paired with the Ledger live app, you have the easiest way to start managing your own private keys and taking control of your Bitcoin.

随着价格飙升，你肯定不想把币留在交易所。我们见识过交易所的结局，对吧？现在你可以通过Ledger Live应用完全透明地发送和签署比特币交易，说实话这操作可能更简单。我强调过无数次，我至今仍在使用2017年购入的同款钱包。

And with the price booming, you don't wanna be leaving that on exchange. We've seen what happens with exchanges. Right? Now you can send and sign Bitcoin transactions with full transparency in the Ledger Live app, And, honestly, it can be easier. I've told you a million times, I'm still using the same wallet I bought back in 2017.

我的Ledger Nano S至今仍在服役。最近Ledger还推出了比特币橙色特别版，并开展了一项超酷的公益活动：每笔销售将捐赠5美元给比特币开发者社区，首期合作对象是我非常看好的Brink组织。想了解更多或购买硬件钱包，请访问Ledger官网shop.ledger.com（即sh0p.ledger.com）。

I've still got that same Ledger Nano S. Also, Ledger has recently released their Bitcoin orange nanos, and They got a very cool initiative going with it. They are giving $5 from every purchase to the Bitcoin developer community, and they're starting out with Brink, which I love to see. So if you wanna find out more, if you wanna get yourself a hardware wallet, please head over to Ledger. That is shop.ledger.com, which is sh0p.ledger.com.

现在说说Swan——他们的使命是帮助数百万人进入比特币世界，并打造了最快速便捷的入门方式。通过Swan应用只需几分钟就能注册并开始购买比特币。向亲友推荐比特币从未如此简单，而且现在Swan平台前1万美元的比特币购买完全免手续费，零成本即可起步。不过别担心，

Alright. Let's talk about Swan and their mission to help millions of people get into Bitcoin, and they've built the fastest and easiest way to get that started. It takes just a matter of minutes to sign up and start buying Bitcoin with the Swan app. So introducing friends and family to Bitcoin has never been easier, and Swan now has zero fees on the first $10,000 of Bitcoin buys, so it is free to get started. But don't worry.

若你已是Swan用户，接下来1万美元的购买同样免费。如果你热衷推荐比特币（我知道你是），选择Swan根本无需犹豫。一次简短对话就能让朋友从零接触比特币，只需让他们在应用商店搜索Swan Bitcoin（拼写为s w a n b I t c 0 I n），下载完全免费，即刻开启旅程。

If you are already a Swan client, your next $10,000 of buys are also free. If you love to recommend Bitcoin, and I know you do, Swan is an absolute no brainer, You can take a friend from zero to Bitcoin in just one quick conversation. Just tell them to search for Swan Bitcoin in the App Store. It really takes no time, and it's completely free for them to get started. Again, search for Swan Bitcoin, which is s w a n b I t c 0 I n, to download the app.

好的。现在聊聊Casa。我们来谈谈冷存储。注意听。看好了。

Okay. Let's talk about Casa. Let's talk about cold storage. Now listen. Look.

过去一年比特币价格暴涨，我知道你们中有些人见证了比特币资产的飙升。或许你还没考虑过多签存储方案，或许冷存储不在你的计划中，但你必须重视起来。我使用Casa服务已近四年，体验堪称完美。

The Bitcoin price has been booming this last year, and I know some of you will have seen your net worth in Bitcoin flying up. You may not have considered multisig storage. You may not have considered cold storage, but you need to get on to this. I have been using Casa for about four years now. Absolutely love it.

感觉舒服多了。现在我的比特币被锁定在一个多重签名方案中，而Casa是最佳选择。但我也想告诉你另一件事。他们推出了一款颠覆市场的产品，解决了我长期以来的担忧——遗产继承问题。我知道有些人通过设计谜题或与他人共享密钥来规划比特币遗产，但Casa完美解决了这个问题。

Feel a lot more comfortable. Now my Bitcoin is locked up in a multisig solution, and Casa is the very best. But I also wanna tell you about something else. They have brought a product to the market that changes everything, something that I've needed, something I've been worried about for a long time, which is inheritance. I know some of you plan your inheritance with your Bitcoin by creating some kind of puzzle or sharing keys with other people, but cars have nailed it.

他们将这一功能原生集成到应用中，你可以在应用内直接管理托管和继承事宜，确保万无一失，确保所爱之人能在你遭遇不测时继承比特币。只需指定受益人，当你离世后，你指定的接管人等待六个月即可接管比特币。Casa还会提供协助，确保比特币在此过渡期间的安全。若想了解更多，或意识到是时候选择优质冷存储方案，请访问casa.io。这就像加拿大不再铸造一分硬币，因为铸造成本超过面值。

They have built this natively into their app so you can handle custody, and you can handle inheritance directly in the app, ensuring there's no mistakes, ensuring that people you love inherit your Bitcoin should something happen to you. All you have to do is choose a recipient, and when you snuff it, when you're gone, the person you have put in charge waits six months, and they take control of your Bitcoin. Casa also provides you with a helping hand to make sure your Bitcoin stays safe during this transition. So if you wanna find out more, if you've realized it is time to get yourself a good cold storage solution, please do go and check out CASA at casa.io, which is casa.io. It's a bit like how in Canada, they they don't make the the penny anymore because it costs more than a penny to make a penny.

对。对。对。然后我们就开始讨论这个话题...

Yeah. Yeah. Yeah. Yeah. And so we started talking about oh, yeah.

我们讨论这个是因为长期以来有个不成文规定：我们在政策中设置了标准性检查，禁止某些操作。比如当我的节点内存池收到创建粉尘输出的交易时，即使共识有效也会拒绝。这原本是为了保护网络资源，阻止可能影响可扩展性的行为。现在人们质问：为什么不能对铭文也这样？毕竟政策可以禁止某些行为。

So we're talking about this because this has been like a very very long standing rule is we had these like standardness checks in policy where you can't you can't do that. Or like if I I'm running my mem, I have my mempool and I'm running my node and I receive a transaction that creates a Dust output, I'm gonna go ahead and reject it, even if it's consensus valid. And that was kind of there as I guess, the idea is to protect network resources to encourage people not to do things that are, you know, gonna hurt the scalability and whatnot. And people are like, why can't you do that for inscriptions? Because like policy can ban or like can prevent, things.

于是人们提议：为何不制定政策规则，当交易中出现类似NFT的铭文时就禁止？这个想法已通过不同形式的拉取请求提交到比特币核心代码库。你知道吗？这些已在GitHub上引发论战——那里不是推特，是讨论代码的地方。

And so people are like, why don't we have a policy rule where if we see an inscription in a transaction, like NFT like thing, then we also ban that in policy. And this this idea has been dressed up in like many different trench coats as different pull requests to the Bitcoin Core repository. Oh, do you know about this? And like they've become like flame wars on on GitHub, which is not Twitter. It's it's a place where you discuss code.

作为维护者...

It's, you know As as maintain

作为维护者，这是你们争论过的问题，还是说你们显然不该介入？这对你们而言是否属于审查行为？

as maintainers has this been something that's been debated or is it very obvious that you just don't get involved in this? This is censorship to you.

我们之前讨论过这个，因为我觉得人们对待拉取请求很认真，对吧？任何人都可以贡献，但这不意味着我们会认真对待每一个请求。当我说'我们'时，仅代表我个人观点。

So we've talked about this because I think peep like, people take poll requests seriously. Right? Like, anyone can contribute. It's not like we're like, you know, like we'll take every poll request serious. I'm when I say we, I speak on behalf of myself and not anyone else of course.

但确实，仔细想想，我其实发布过一份很长的文档，总结了这类提案的某个迭代版本。因为它收到了数百条来自不同人的评论，专业程度参差不齐。所以我非常认真地尝试索引每一条能找到的评论，归纳论点并权衡技术论据的优劣。是的，我非常认真地审阅了那个PR。

But yeah, like, you know, you think about it. And actually I published a a like really long document kind of summarizing one of the iterations of this kind of proposal, because it had like hundreds of comments from various people, being, you know, to various degrees of professionalism. Yeah. So I like tried very, seriously to like index every single comment I could find on it, like summarize the arguments and weigh kind of the the merits of the technical arguments against each other. So I I yes, I did very seriously like review that PR.

如果你需要，我可以把文档发给你。我们之前讨论的是——为什么不能对铭文做同样处理？我认为有两个原因：一是你可以在自己的节点上实现。

And I I can send you the the doc if you want. But like what we were talking oh, so the idea was like, oh, why can't you do this for inscriptions? Right? And like, I think there's there's two things. One is like, you can do this on your node.

对吧？但关键在于，'我想在自己的节点上这么做'和'希望比特币核心默认集成这个功能'是两回事。我想强调两点：一是比特币核心没有直接审查内容的能力。

Right? But like when you're asking, like there's a difference between like, I wanna do this on my node and I want Bitcoin Core to ship this as the default. Right? And so there's two things that I really just want to put out there. One is like Bitcoin Core doesn't have the ability to just censor things.

正如我所说，我们只能发布易于运行且激励兼容的软件。如果我们开发个软件然后对矿工说：'能运行这个吗？虽然你会损失30比特币手续费...'他们肯定会拒绝。

Right? We can only, like like I said, we try to publish software that is accessible to run, but also incentive compatible to run. If I publish like, sorry, not I, but like if we create a piece of software and we're like, hey miners, can you please run this? You'll lose like 30 bitcoins in fees, but like, please run this. They'd be like, no.