Bitcoin can scale to millions of transactions per second by implementing multicast networking, which enables one-to-many transaction distribution across the globe within milliseconds, overcoming the bandwidth limitations of traditional peer-to-peer gossip protocols that become inefficient at high throughput levels.
Install our extension to search inside any video instantly.
How Bitcoin Scales Past Millions of Transactions per Second | Jeff Harris
Added:Hello and welcome to another episode of the Proof of Claim podcast. I'm your host Todd Price. I'm joined with my co-host Anil Barrett and today we've got Jeff Harris on. So we were just trying to get an explanation of what Jeff is to to introduce him in his edification, but he's a a veteran of computers. He's been in the game for over 30 years in um all kinds of distributed applications, distributed systems, network engineering. Um so he's kind of touched every point of the stack at some point throughout his career. I met Jeff um at the London blockchain conference in 2023. We shared a couple of days together manning the the BSVA booth and um we had a great great set of conversations just going deep into the technology and it was a real real privilege to get to speak to someone who'd had the intimate experience building out Teranode um understanding how the system worked understanding how it interfaced with the broader network of the internet and being able to address any questions at any level. So, um I'm hoping today we can really get deep into some of the technicalities of Keranode and the future directions that we're going to have to take um the Bitcoin in to to scale it to a planetary level. So, welcome Jeff. It's great to have you here.
>> Thank you. Thank you for having me.
Yeah, >> I've been um following you on on X and LinkedIn and just seeing that you've been um pumping out a bunch of code related to um you know what your claims are. This is how we get Bitcoin scaling past, you know, the single digit millions into, you know, several orders of magnitude above. So, I'm keen to get you on here to >> prove your claim or just at least sub substantiate it and let's see how we where we we take this to um learn about what the future looks like for these kind of blockchain systems and you know what what you're aware of in terms of you know the feasibility for scaling Bitcoin as well as the feasibility of scaling other blockchain systems and you know how how does that looking for you?
So yeah, welcome and um let's let's get into it. So do you want to share a little bit with what you've been recently working on with um your Bitcoin multiccast sharding proxy or [laughter] whatever the label was on?
>> Indeed. It's it's actually evolved into kind of a network suite. Um the there's a bunch of moving parts. The main thing is that we need to be able to distribute transactions efficiently across the planet um in a fair manner and it needs to do so in a scalable way and it requires uh in order to meet that scale it requires a substantial investment and and that means standing up a real network to make it happen. Uh the way that Bitcoin has scaled thus far is through various peer-to-peer public internet um type structures. The current structure is a lib P2P gossipbased uh distribution mechanism. Basically lightweight announcements from nodes get sent out and then many other hosts all download from that node. Um, and it ends up being a little bit tricky because of the amplification of bandwidth that starts to occur at at certain throughputs. Um, it's kind of my supposition that about 3 million, maybe four million transactions per second is actually going to start running into problems because most hosting providers have 5 gigabit, 10 gigabit flow caps uh for for a single server. Um, you know, 10 gigabit internet is pretty standard these days. 25 gigabit is even more, but bandwidth is still a a darn premium. Um, and hosting providers tend to treat it like almost as a a capture mechanism, right?
Um, for instance, AWS makes it very it they make it free to get data onto their network and then they make it expensive to get it out. Other hosting providers have their own types of uh games that they play. You know, uh they they'll give you up to maybe three gigabits, but then to push up to five gigabits, you're spending $800 more a month, right? Um there's just all kinds of uh funnels that push you into higher spend usually.
And that's just the hosting game. It's just the way the the game has evolved.
And so Bitcoin in order to break out of the science project um status and to get to real commercial stability, it it needs to be professionalized and on a public internet level that is data centers, collocation, internet exchange points, transit uh not even not transit but peer-to-peer direct connections data centers um larger lengths than 10 gigabit, you know, 100 gigabit, 400 gigabit. Uh we've got 1.6 gigabit uh or 1.6 terabit network interfaces now that they're testing. Um >> and the fiber is still still there and still viable. You just change the interfaces and you're you're good to go.
Um Yeah, >> the the network scales the most. I don't know if you've ever seen charts of Moors law put next, you know, overlaid on top of where memory scales, um, storage, etc. Bandwidth actually scales the most.
Network network always has just because the innovation with light is is more pronounced. So, >> yeah. that's less um bound by the material manipulation. You're essentially just optimizing how the the photons are arranged in a in a tube or a fiber or something like that.
>> Yeah.
>> Being able to write to the to the frame.
>> Yeah. And the and the small world network requires very dense connections between the the hosts, right? The the miners. And uh the faster that happens, the more money they can make uh the you know less latency uh more more transactions, more connections, you know, uh it it the network becomes a utility and you know teranode is kind of an engine right car but it needs a track and the network's the track.
>> Yeah.
Yeah, it seems, you know, there's there's a few operators running it now, but um you know, trying to trying to put a million transactions at them, you you you I don't have the the highway to to put that much traffic to it. like my laptop, I've maxed it out, sending 15,000 transactions a second to arcade, but then my the the router from my home's not going to be able to let me do that, send send more than that really.
Or maybe it would, >> but actually I mean 10 gigabit uh consumer based fiber is actually a thing in certain places. I don't have it, but other people do. And you know, I think you can get a million per second across two gigabit. That's that's no problem.
Um >> Yeah. Right.
>> You know, um there's also you have to think about the transaction flow because it's supposed to come from users, right?
It doesn't come from miners. Miners just accept transactions in and then they they verify them and they share them with others, right? Um, so what I Okay, so the source material comes from a person that everybody knows and loves and he he wrote very deeply on the the subject of how to actually scale the the core of the miners. And so my implementation is based on that design.
And in this design, it breaks apart the ingress users from the core consumers, which are the miners. On top of that, it's services overlays.
On top of that, it's it's users, right?
Um, if you think about what multiccast is, it's just think about it from the standpoint of one transaction. One transaction.
We're putting it on the wire. It goes to every party across the planet at once within milliseconds. That's the that's the goal, right? To create a multiccast fabric that actually traverses the entire globe to consumers that actually need it. So, >> what what is like I I know, but I think many people don't know necessarily what multiccast is. Um, >> that's that's exactly what I just described. It's the it's it's the ability to put a network packet onto the wire and then subscribers anywhere else on the network, >> receive that transmission all at once.
>> And that's different than a uniccast arrangement >> because a uniccast is one to one, right?
And a multiccast is one to many.
So if you're having before you mentioned lib P2P um as a a gossip. So, if you're you're putting a packet on the wire um under gossip, you're gonna then and you've got 10 peers, >> it's it's going out and and the if the packet is, you know, whatever 220 bytes for a transaction, it's going out as to, you know, be 2,200 bytes, but on the >> and and it actually doesn't even work that way now. Um, since we have Teranode, the old peerto-peer worked kind of like that.
Excuse me. The new peerto-peer, uh, Teranode actually announces sub trees.
>> And when they announced the they announced the subtree availability to the gossip network, the gossip network's made up of of miners, those running Terod and other peers, right? uh those >> yeah those peers uh reach out to that terode's asset server because there's a there's a data URL inside that announcement. So they they go out and they they talk directly to the terode and pull that data right if they're if they pull the subree data if they're missing any of the transactions then they pull the transactions from them too. So already it's a it becomes an exponential bandwidth problem the more participants there are and uh honestly it's never been tested over the public internet.
Um, we did test it at a million plus per second across the internet, but it was writing on AWS's network. And so, um, and it actually used VPC peering so that we wrote across AWS's actual um, actual network links. It never left their network. So, we've never tested it on separate public internet connections. I think it'll probably be fine for a million, two million, but like I said, depending upon the hosting provider, at some point you're going to those flows are going to start running into bandwidth caps and it might get pretty expensive really fast. And the other thing about this is every minor then has to run their own network operations. They've all got to decide that okay well we've got a lot for so much bandwidth and you know we have all these people pulling all this stuff and you know and it's it it changes you know uh there might be periods of lots of usage there might be periods of not a lot it it becomes a another operations expense right for these these these miners that they really their business is validating transactions, creating blocks.
That's that's their business.
>> Emanating disseminating the ones they've received to others.
>> Yeah. So, what we want to do, what I want to do is to create many ingress points all across the globe.
You send your transaction to a standard DNS address, standard IP address.
Anywhere you are on the planet, it goes to the nearest ingress point. It's the multiccast fabric. The multiccast fabric transmits it everywhere across the globe directly to teranode.
I want to frontr run the lib P2P pull process and I want to replace it. I want that there in case there's uh it needs fallback, right? In case um in in case they didn't receive a a multiccast transmission for whatever reason. Um which is very unlikely the way that the network is actually designed, by the way. We get into that. Um >> your network that it's designed or the multiccast >> the multiccast. Yeah, the multi the way that the multiccast um frame format works and the ability to retransmit um and actually fill gaps if it's it's like magic. It's we'll get there. We'll get there.
Anyway, for now, what I'm saying is is that lib P2P gives a um it gives asurances up until a certain throughput level, a certain threshold, right? Beyond that, it's not really tested at super high throughput.
So any anything can go and the math that I've seen shows that it starts to break down for various reasons.
>> Um >> so this is just the next step.
>> It's just the the professional network.
That's all that that's all that this is.
But we have to do it in a way that's smart. Mhm. And so you you got the you know some of the the brief from um that that the article that Craig published.
>> Yeah.
>> Like how much of that was you know revolutionary to you or novel to you or like was it was it or did you already know most of that and it just for helped you formalize it or >> absolutely not. So, the article that we're talking about was written in September of last year, 2025. Uh, it was called Multiccast in Multiccast and and AnCast, I think, was in there also. I don't know. It was a long title.
>> Anyway, uh I started reading that article in January of this year. I took a I took a little bit of break after the Teranode project and uh rested and relaxed and then started getting bored. So I started reading in January. Uh it actually took me a while to to understand it. Um I had to read slowly and digest. But as I started experimenting with some of the ideas in it, I started to realize that, oh, this might actually be more doable than I I first realized because you got to understand in 2023 when we first started looking at multiccast, when Jake Jones and I were looking at it at BSVA, no one had any clue. like I knew how to put together a a uniccast routing network and then standup PIM on top of it. Um I had never done it before but I figured it out and we got actual multiccast working but it wouldn't work across AWS just for various reasons. So um so we kind of abandoned that because we really wanted to hit our million TPS goals and we did a couple months later.
Uh so we really I didn't really even think about multiccast. um the rest of the time that I was working on tyrannode. Um [clears throat] so this year January I picked this up uh the the biggest thing to understand for me was how to do the actual sharding of the transaction flows. So the idea is that we've got a billion transactions per second, right? You can't process a billion per second on one computer.
doesn't work. Uh, and you can't flow that many over one network connection really because it's terabytes of data, right?
>> Maybe someday, but not today.
>> So, what you have to do is you have to break this up into in a deterministic way so that it can ride multiple lanes at once, right? So, think about it as a highway that can take so many cars. And in order to alleviate this, well, you just stand up a new one right next to it. And the cars that are going down these lanes, they all conform to a certain address, right?
A certain portion of the SHA 256 or the U, sorry, the the TX ID address space, right? or the the the actual SHA 256.
[clears throat] So you take the the big Indian version of that and you do a little bit of fancy bit shift math.
That's all you got to do. And you split it deterministically into groups. And so you decide, okay, well, I've got I'm going I'm shard bits zero is all in one group. Shard bits one is two groups.
shard bits two four groups. The magic is that those group names never change. So if you've got 10, let's just say you have 10 subscribers on the network that all want to see, they all are expecting to see traffic on this one group and then you shard it into two. Well, all they have to do is join the new group and then suddenly they've got the entire flow.
And you can keep doing this. Uh, I've designed it to where you can split it into 4,96 different lanes like that. But can you imagine >> can you imagine managing 4,000 network links like that all had different group subscriptions on them because they all had a different piece? I think that's way too many, but it's it is what it is.
It's it's there. Um because I think it some you know as time goes on network interfaces get faster right the more bandwidth right so you're always going to there's always a a a pushpull between those two things so you can you can shard more or you just up upgrade the links either way. So the core of the network the one the part that's passing a billion per second can be split up into different groups. So you could split that up into 12 different groups >> of uh of 100 million each, right? You've got 1.2 1.2 billion total.
>> And you're essentially like indexing the instead of getting the state explosion of multiccast, you're just indexing by the leading bits of the TXD.
>> Yeah, that's for the transactions. Yes.
Um so basically when they hit the proxy, the proxy makes a decision. It it says it looks at the TXD really fast and it says, "Okay, which group does it belong in? Put it here. Pass it on through." It also does some some stamping which is a it's stamped with a composite hash key that's made up of the source ID the source IP address the group index that it's on and also the subree ID that's added to the transaction as well. So there's three dimensions there. And then next to that there's a monotonic sequence number. Um that's I think it goes up to four billion. So as the proxy is sending these packets out, it's stamping them with a key that can be used to reook up that frame. Um people that are watching probably don't know but uh multiccast is not a it it can't guarantee delivery.
Uh first off it uses the UDP um packet type which doesn't have any provisions for retrans retransmission like uh TCP does. TCP is a reliable protocol and UDP is not. Uh so when you put a packet on a multiccast network, there's no way to guarantee that everybody has got it except for your your network convergence skills uh and reliability thereof. And so >> and so you need a way for receivers to identify if they've missed certain transactions. So, think about a transaction comes in in Japan and somebody in California uh is trying to listen for this entire transaction stream and one gets missed.
Well, what can you do? Well, you look at the sequence. You look at the sequence of the flow and you realize that you've got packet number 10 and packet number 12, but number 11's missing. So you take that the composite hash key >> and sequence number and you request a retransmission and because it's on the multiccast network you can have many ret-ransmission points and all they have to do is cache the entire flow of transaction data over a period a time window of like maybe a minute maybe 10 minutes I don't know it depends on how granular you want to get and at at a billion per second it starts to the there's a lot of keys there, but it's just RAM. It's just RAM and and servers.
And so all you've got to do is load balance the incoming retransmission requests. If you have a reliable network, you never need a retransmission request.
So, so what happens is you've got transaction comes in ingress gets put on the multiccast network. It goes end to end to n number of consumers miners >> at that at that edge >> the before it gets delivered to the minor over uniccast TCP so that it's reliable >> it just it just repairs any gaps. So if you have a network cut, if you have a um if you have an outage of a router, the thing can actually heal itself because all you have to do is a second path, do a retransmit request, you've just filled all the gaps. So it's a this isn't novel because the the Navy actually has a a protocol. what's called norm uh knack oriented reliable multiccast protocol and theirs is works a little differently because it's kind of block oriented and it uh it does some optimizations for like uh retrieving missed pieces for other consumers also but we don't need any of that because we have a transactions are just an atomic unit right it's just send the transaction through, just put it out on the end. You know, we don't even need encryption. We don't need anything because Bitcoin's public, right? It's literally just delivering them. It It's like a It's like a reverse post office, you know? Uh you send your transaction for free everywhere and then you uh miners pay to receive it. That's the idea.
So that's that's that's your sort of route to commercialization is by you giving them that reliability and and an efficient way of ingressing that you know >> and there's really and there's really no other way to do it that I've that I can think of because as this thing scales you need those direct connect um data center port connections you know because like I said over the internet flows start to break now. So it it turns into private connections between miners. But this allows everyone this allows all the miners to streamline their network ingress to one flat rate which is the entire rate of the network the the the the overall TPS of the network. So rather than potentially having bandwidth spikes because of that issue we were talking about with lib P2P and um other miners suddenly hitting them for information.
You just deliver it steady state to all continuously.
That's the idea.
And so what you're trying to do is like, you know, create a system that scales unboundedly here that you can just sort of commoditize a reliable delivery of that to everyone and you just have this sort of fabric that they can count on and they they um you know, you need to need to come to you because you're the one that's that's >> well I've also designed it for competition so it works interdommain. Um this isn't uh it isn't a proprietary solution. So all the code is open source. All the standards are open source. They're designed for uh source specific multiccast which actually works across interdommain links. The problem is is that no ISP on the on earth well not on earth but no major tier one ISP tier 2 ISP accepts multi-protocol multiccast across their transit links.
Right. They just it stops at the ASN boundary because it's been considered a kind of a security issue honestly because of DOS and all kinds of other fun things. Um so my solution is designed to be interdomain. So if somebody else wanted to stand up their own network, we could peer together. We could have ingress points and miners on both sides. It worked just fine. the solution works just fine, but somebody's got to stand this thing up. And as far as I'm aware, there's not anyone else in the BSV space that's focused on the data center and the network and anything above scaling teranode itself if right >> I'm not aware of >> maybe maybe a couple of there might have been a spirious claim from someone [clears throat] about >> um >> actually Tai at Babage did some work thinking about multiccast uh and I've read his work several times actually um and I like some of his ideas on resource configuration um resource selection I guess is a better term for it uh but it didn't it didn't really go far into actually spelling out how to build a network right how to architect it for scale and the source material I used did Well, >> um, >> what does that say to you that that you know, you've got a 30-year career in this sort of stuff and obviously you know what you're speaking talking on and then you read an article by someone else that >> kind of gave you a whole lot more.
>> It's it's divine inspiration. That's that's what I'd say because, you know, it came at the perfect time. uh I needed a new project to pivot to and this one sounds really cool. So um >> from your capacity to sort of discern that does does does the individual that published the article have a extreme you know knowledge of these kind of systems and networks and such. Well, I I I tell you what, the implementation details weren't there.
>> It was just a design. It's just a design. I had to figure out how to build it.
>> So, it's been sweat equity in this, you know.
>> Yeah.
>> But it comes because I've built networks, routers, switches. Uh I've I've did a lot of Cisco certifications decades ago. Uh, I was running an ISP in 1995. So, I wasn't running it. I was a junior cis admin. I'm not going to get ahead of myself there, but I was a seven I was a 17-year-old junior citizen cisadmin, right? Um, so I'm I'm not a computer science background, not academic. I'm I've always been a working engineer and I've just kind of worked my way up. I've worked at startups. have worked at enterprise, retail, government, like all kinds of ISP, service providers, um all kinds of stuff. So, I've had a lot of cool experiences >> and we kind of jumped, you know, straight into the weeds then, but um >> we did >> just on the little precall um you know, Anil asked you how did you get into Bitcoin? What when when you got into it?
And so that that's um what what you shared was a very early early um entry into it. And how much has um you know what has it been 15 years or something 16 years since then? Um how much Bitcoin occupied your passion at the same time as you know your experience in network engineering and system engineering?
>> Yeah actually um good question. Uh so just to Just to relate the story, uh, found Bitcoin late 2010. By early 2011, I was mining it on GPUs because that you could still do it for $100, you know, get a nice Radeon card. Uh, [clears throat] I was along for the ride all the way up through the man, I remember the the first My Bitcoin scam or whatever. I'm thankful that I didn't actually get I didn't actually lose anything in in scams and exchanges going bust and whatnot. I had a good friend who lost a bunch on empty gauss actually. Um, and it sucks because I I knew something was going to happen and I just I didn't tell him because I don't know why I kind of kicked myself that I didn't uh I didn't intervene but I I lost touch with him. So, I don't know what I don't know if he ever got made whole or whatever. You know, I think people waited forever about that.
Anyway, >> I think I think that people did get made whole uh more like maybe two, three years ago or something like that. Yeah.
>> Yeah. Yeah. He um Yeah, that was it was unfortunate. Anyway, uh so around 2013, I actually started working for a large managed hosting provider, one of the big at the time it was the fourth largest cloud um company called Rackspace Hosting. Um, and I founded the Bitcoin and well, I called it cryptocurrency group because back then I didn't understand the difference between uh between Bitcoin and cryptocurrency.
>> Well, what how would you define the difference then just for >> Well, cryptocurrency means secret money and Bitcoin is not secret. Bitcoin's public.
>> Yeah. Yeah.
>> And also it's cash, not money, right?
Money is a different uh money is a different legal definition. It has different connotations, right? So, what we're what we're working with here is electronic peer-to-peer cash. And >> yeah, >> um and so I I should have called it the Bitcoin Group, but by 2013, the the lines were murky, let's just say.
>> Uh so I uh I was always a a Bitcoin advocate. Um I uh uh encouraged other people to contribute as much as possible. Uh the thing is is I never really uh waited deep into the waters with Bitcoin. Um professionally, commercially, hobby, you know, I helped I helped a guy uh recover some lost pass pass wallet passphrase or whatever, you know, did some odd things like that. Uh, but I just uh I never really worked with it like that until I ran across the Unbounded Enterprise guys way back uh way what was that like 2016 2017 whenever that was. Um kind of worked with them for a year or two. Um really that didn't really work out unfortunately for them. I think they got caught up in some um I think they got caught up in the Kristen Agar Hansen affair unfortunately um because they were one of the businesses that Inch Chain went well while he was CEO did some deals or something. I'm not I'm not privy to any of that stuff. I don't know how it went down but >> pretty one-sided. Um, >> yeah, I can only imagine that that actually contributed to asset layer failing afterwards. So, not sure. That sucks for Jackson Laskkey because he's a really great guy. Uh, and Dave Mullen Mirror also. I really like those guys.
Uh anyway, did did some of that work and then uh some other things and then fast forward to 2023, I'm working for BSVA, working on Teranode.
Uh actually doing my actual dream job.
Um, I got recruited in from a friend of mine who I worked with at Rackspace actually because he went to AWS and wound up being our uh uh >> man on the inside.
>> Yeah, man on the inside. Not actually officially. He uh he dedicated his time to Terry because he believed in it.
Yeah. He didn't get paid. I don't think he really got uh credit for it, you know, as his job at AWS. Like it it uh he really stuck himself out believing in the project. So, I gotta give him he knows who he is. So, uh got to give him some props. I spoke to him last night and he said that you guys are in regular regularly chat and um >> it sounds like he's getting a little bit of traction now with um customers who are interested in using the tech and he's got and some prototypes and he's saying yeah what >> how do we still what is the um you know the coherent suite of tools that can offer and um you know I was sharing some some of the stuff I've been working on with with him and in particular, he had an IoT customer and I said, "Oh, check out this little edge kit I hacked together, you know, see see if there's anything of value in there." And um, >> you know, I'm keen to, you know, cross-pollinate and hope hope that something crystallizes that just um, you know, delivers a killer killer solution to some particular use case. And you know if you get a couple of those wins um then a whole like IoT in particular I think that industry um you know if some if I was down at the le e conference a couple of weeks ago and sitting next to a professor associate professor from Japan and he was saying they were trying to do sharded lightning network IoT stuff and you know they're just publishing their negative findings and I said you know you can do you know 100k TPS for fractions of a cent And he he was like re really I'm like yeah L1 I'm very interested by that like I've got smart factory smart city you know aspirations that we want to do as a research project and so that that was my kind of inspiration to sort of formalize some of the little tinkering I've been doing on the ESP32s and that and wrote up a a little edge kit for him and said have a have a have a look at this see if it makes sense. you probably know the you know the issues and what you the you know the requirements and such but maybe this is a a bit of a head start to to look at it. So um yeah shared that with with um our other mutual mutual acquaintance and um yeah hoping someone can you know deliver something.
>> Yeah, I'm I'm glad he's still uh still at it. Um yeah, I'm glad we all are.
Honestly, >> we're we're like the uh uh you know, it's whittling down to a core group of hardcore enthusiasts.
>> Uh >> I got the bar. I can't really imagine, you know, it's the hard path, but I can't really imagine pivoting to go anywhere else because it's just like this is, you know, when you have the conversations having with you, you know, there's very few people I can find. But when you when you do it, you're like, >> where where else where else would I rather be than at the the tip of the spear of scaling, you know, money, peerto-peer cash, and and data a data plane for, you know, you can just see the relevance to to every every vertical and you see the inefficiency. You see the the value lost in fraud and you know data data l a lack of data integrity and a lack of being able to trustlessly coordinate between you know not necessarily like adversarial but non-g guaranteed to be you know always cooperative you know lossless channels and such and that that's the solution for it but um yeah it's most people don't You know, if you talk about data integrity, they don't it makes their eyes glaze over.
>> What is that?
>> Yeah. Or you talk about dig a digital identity and they're like, "Oh, yeah. I just log into my account." And you >> and and I have to be honest like I'm not uh I'm not the one that's I'm not deep into script engineering uh decentralized identity solutions um wallets none of that you know my uh I mean I okay I try and keep a breast of of all the BRC's that I can >> [snorts] >> u but I've got my own that I'm working on and I'm kind of working on a different layer of the stack than I think Even all the other Teranode engineers were uh which is the actual infrastructure where the systems are the operating system the the network the storage um all of these uh the clusters um all of these lower level solutions that are required in order to run the thing right um but also scale it so uh I find that it's a very lonely existence Because honestly, I've been working on this project for seven, we're in I'm in my seventh month if I count, you know, studying for the first month or so and then starting to write code. Um, you know, six, seven months in and I still haven't had really good peer review or, you know, people telling me, well, you you're doing this wrong. Uh, you you know, I'm kind of operating in a bit of a vacuum. Um, I try and share things with people as I can. Um, but I think a lot of the focus at BSVA in particular has just been on getting Teranode running, getting it getting it up and live, getting SV node off the network, you know, or at least building everything so that that can happen. Um, and so I just think that a lot of people haven't had time to really think about this, but I feel like >> we I know everyone wants to know where the throughput's going to come from, where are the transactions actually going to come from, what are the solutions, right?
I have a feeling that well I know for a fact like in my head I've got certain sources of transaction entropy that I know could drive throughput through the roof. But getting to that point, you know, commercial solutions is uh is another story. But I think a lot of it is going to come down to plumbing stuff that's on the back end that no one really thinks about because Bitcoin can just do it easier or more more uh more accurately, more transparently, not even transparently, more accountably. We'll just say that. Um >> and I and I really think this real just to close that out, I really think the solutions are actually going to come from science. high performance computing uh academia type work first and I think that's going to get the throughput numbers up there into the millions and then either we're going to get eyeballs or we've got to well a we've got to figure out how to scale past that for sure because you can't stop right scale or die [laughter] And that's you can't stop at one million. You gota just you got to keep going. So, >> but also there is that like I we were just talking about this with Connor the last week, right? Like there is a there is an actual time ticking time bomb that we're we're facing. We have to actually get enough transactions on the network to >> n as long as someone's willing to expend the the power output for hashing gear and to to maintain it. You know, honestly, this thing, this coin could be a dollar and I bet I bet somebody would believe in it enough to spend that to keep spending that money. I mean, if it were if it came down to miners had to sell Bitcoin to survive, >> then yeah, [clears throat] then I'd agree with you. But right now, we've got certain people that really believe in this project and they're >> I don't believe that they're going to uh leave us at the mercy of coin price economics. They haven't already, you know, I mean, we're we're still here. We still have a viable chain even even through all the the wars, the hash warriors, etc., you know, even with it only averaging what is it like 40 something transactions per second now.
You know, I know we've had a lot in the past.
>> Yeah.
>> But honestly, right now, okay, think about this. We've had we've topped SV node out at 13,000 transactions per second. We've kicked apps offline. We've kicked nodes offline. You know, we've caused all kinds of havoc anytime anyone tries to blast transactions out there.
No matter if it's a legitimate app or if it's testing, spam, whatever you want to call it, >> you know, it's not spam. If it's a paid transaction, it's legitimate. Anyway, uh it makes no sense to expend any revenue on that right now. What makes the most sense is to spend the money on scaling the software so that you don't run into those roadblocks, right?
>> And that's exactly what's been happening the last two years. We're getting really close. I think by the end of the year I hope I hope we've got SV node off the network so that we can start hitting the the terodess that are on there with a lot more and it doesn't take that many Bitcoin to run a million transactions per second continuously right and a lot of that goes right back to the miners which might have a incentive to do more like there's there's um you know, Bitcoin's got its own economics. When you start to add actual throughput on top of it, it changes everything. Um I would agree with you though, the having is is a I watched that episode, by the way, just to just let you know, I watched I watched it yesterday and the day before and I do agree with you that the having is coming up, but I don't think the economics solely depend on the the coin subsidy right now. Um even in four years you know when it's what is it like one something or whatever uh >> I think >> or six years whatever it is you know when it gets under one you know >> yeah 0 point something would be 0.7 I think would be the second having after >> as long as as long as the miners are being paid in transaction throughput no problem >> but that if they're not >> No that's what I'm saying though right like I I realize is we we need to get enough transaction volume because the the the h this block reward uh subsidy is going to be so little.
>> Yes.
>> So like that's why I'm saying we have a ticking time.
>> There's no incentive to actually mine it. Yes. I agree. So the incentive actually has to come from somewhere else. So in the absence of throughput and the absence of coin price >> Yeah.
>> the incentive has to come from somewhere else. And that's going to be the goodness of somebody's heart. [laughter] God, >> who wants to keep BSV alive? They have.
They have so far. I mean, there's a lot of people that are really invested in this project.
>> I get that, but we can't really on that.
>> No, we can't rely on that. So the push is >> get the thing scalable, get the throughput figures there, and like I said, if there's a research project, if there's a university project or something that's funded to to do some kind of scientific thing that you can't you've never been able to do before because we've never had a a distributed transactional system like this.
>> Yeah.
>> Who knows? And you know, we used to have tons and tons of traffic on BTC. Um, I remember being on the old Bitcoin Cash Slack when all of that transaction traffic just stopped. And I remember uh because uh our our friend Dr. Wright >> Yeah.
>> was on that chat, too. And he was kind of it was like nod nod wink wink, you know. I don't I don't know what's going on, you know. Uh meanwhile, you know, the traffic had just been going up like this and then all of a sudden it just started >> and yeah, it was doing nothing. And I'd never seen that before in like 10 years, you know, of of Bitcoin whenever whatever that was, however that what 20 something like that >> 2017. Yeah. 2017 2018 was right during that that time, you know. Anyway, we used to have research research transactions.
>> We're going to have those again. I know it. I just I feel it in my bones. I know it's a very old fashioned way to think, but that's just way I you know, there's no reason to spend all of this time and effort to build something like Terranode if that's not coming.
That's that's just where that's just where it at. I don't know of any secret plans. I don't know anything like that.
I just know I just know Pteranod's real.
I know there's a team of like, you know, a dozen or more people working on it and that, you know, I was putting in 60 >> 60 hours a week plus >> uh for two years on that. So, I mean, we were we were hitting it hard and it works. Um >> there's got to be there like maybe if you don't know, but you know, think about it. you doing 60 plus hours a week for two years plus 10 12 other engineers on Terran node doing full-time hours or more that they're not just doing it for like right like you're talking about um what's his name um Calvinire basically like um >> well it was BSVA let's just let's just be specific BSVA and the uh the the the >> the people that run BSVA which I I'm not sure of the entire structure because you know it's a Swiss nonprofit. It's got its own got its own thing but the the we'll just say the BSVA that's their undertaking.
>> Right. So Right. Okay. Got it. Got it.
So it's not actually um >> Go ahead. Sorry.
>> No, you got a >> Well, I was just gonna say that that has to for Is that the right word? Foreshadow basically demand. Like no one's gonna build a a freaking um you know seven 20 30 uh what's what's what's a good analogy here? Nobody's gonna build a a a stadium if they don't expect to have um >> Oh, you'd you'd be surprised. We have something in in North Texas in the I So I was born in the Dallas area and uh still spend lots of time there, lots of family there. And uh when they want to build a new freeway, >> yeah, >> they'll just build the thing. And you know, they don't they don't I'm sure they do traffic forecasting and >> whatnot, but they have to build it first and then the traffic comes later.
>> So sometimes you need to >> you need to build infrastructure in order to prove >> in order to build the things that come next, right? Um, every app so far that's been built on BSV, it has suffered from lack of scalability. The architecture on most everything that's been made prior to the last year, maybe two, um, has failed on a lot of these projects. A lot of lot of wallets, you know, um, you know, it's not everything. Um, but we didn't even have the right SPV plumbing until maybe a year, maybe two years ago, right? Um, and and you need that in order to scale a certain way on at the user level, the the user and the app level, right? The core is even more critical because that that's that sets you up for everything that comes after, right? everything outside of that on the mandala, right? The the core network. Um it you have to get that right. And honestly, if the first Teranode effort would have been successful, um there was a there was a hype cycle that we were all riding in crypto then that could have looked different, you know, if you would have had some real engineering happening back then. Um, I don't I wasn't around then. I don't know what actually took place. You know, I know there's lots of drama, etc. But, >> um, you know, if it would have worked back then, I think, uh, we might be in a different place. We might actually have some of those scale blasts, but we lost a lot of momentum >> over time. And, uh, the new the new Teranode team is topnotch. I really liked working for Siggy. Really have a lot of respect for those guys. I had a lot of fun actually cuz you know I got to bring my infrastructure work side to the to the table you know and um tried to be a leader on the team as much as I could. Um, >> how much of um, you know, [snorts] I don't know if this is going to sound like a weird question, but like how much of your experience with like what you did before getting into Bitcoin and Teranode is different and like and then if even Bitcoin is something different than Teranote, I don't know if that even makes sense, but like how much of um, Bitcoin and Teranote is different from like the work you did What's >> actually it it's not too much different because you know at the at the heart of it the new terode is just a microservices architecture. Um I actually didn't have that much experience with Kubernetes before I joined the team. So I had to really ramp up on that and I'm still not au Kubernetes expert by any stretch of the imagination. Uh but I had to learn that in order to learn how the orchestration works, you know, because all these services have to talk to each other, right? Um and so there was a there was definitely a learning curve there, >> but the infrastructure work from say optimizing and tuning Linux, um uh specking out the right hardware, um network issues, right? network bandwidth routing when routing would break. That's was always my thing to fix, you know, that that kind of thing. Uh uh a lot of that stuff I had already been doing for decades prior. And I actually have experience in very big data centers with very large deployments. Um hundreds of thousands of servers, you know. Um uh I've just done I've done a lot of cool stuff. I always just wanted to work with really awesome network tools. Um I I kind of became an architect over time because I just knew I I know how to stand up scalable operational environments, you know, and it's not just it's not just about building it, but you've got to maintain it also, you know, you've got to make it to where it actually meets those five nines if you if you're building for that, right? And there's a lot that goes into that. I >> I guess another thing is like as far as like where does demand come from is I'd imagine like does any other blockchain project have this type of engineering in their pro you know already and so when demand comes into the entire let's say space and then things start to shake under pressure and stuff like I imagine that you know all eventually gets redirected towards like uh uh you know BSV eventually maybe if they know about us but maybe if they don't.
>> Uh the thing is is that BSV's core strengths are payments right and if you can if you can architect a new way of doing payments that make people money that's what you that's what we need to go after and >> I'm kind of focused beyond the the multiccast network itself because that's just a thing to get us to the thing. Um, I don't know if you've ever seen Halt and Catch Fire, but that's a it was a cool TV series about the origins of personal computing. Y'all should check it out. It's really good. But anyway, inside there, the the guy says, "This is the thing that gets us to the thing, right?" And and that's what I feel like the network actually is because once we have something like that stood up, >> Yeah. we can start to augment apps in certain ways that they are distributed widely and we start to have these new primitives that we've never had before.
Um, you know, being able to create a command and control network that works across the planet no matter where you are very cheaply. Um, it's huge. And the thing is militaries need this, governments need this. Um there there are so many use cases that what we need to focus on is getting the plumbing right, getting the education there on top of it, and then um sparking imagination in engineers minds. And I love all the the app based development that we've got going on.
>> Yeah. Um, and I know some people I know that was a topic on the episode with uh with Connor was that uh the enterprise versus uh startup app, you know, small time apps and whatnot. I think we need both because the the enterprise solutions, they take longer, but a lot of times they're more baked in, right? And payments are a business decision automatically really. I mean I guess you could accept payments personto person whatever you know but that only scales so much right because of various reasons legal etc. So it it naturally trends towards business which doesn't have to be enterprise but it it certainly heads that direction when you start doing any kind of volume.
So, um, a lot of the things that I'm focused on beyond the network are networkborn problems that can be solved with Bitcoin. And the network is the largest source of what I identify as potential payment entropy on the planet.
Um well yeah I'm speaking with some other other um associates or whatever and we're looking at looking at network looking at hyperscalers you know data centers and saying where does this you know fit and I'm kind of saying like the the north star that if we want to think about a product that slots in there is um turning turning the network into commodity type infrastructure that's paid for pro pro rider. So you just pay for what you consume and you create spot markets for the utilization of computer resources. So the more of an infrastructure layer you are, the more you're delivering that computational, you know, resource or um storage or bandwidth or whatever at saturation of the the plane because if you've got spare capabilities in your bandwidth, why don't you sell that as a surplus? If you've got spare storage, why don't you sell that? If you've got spare cycles on your graphics card, why don't you sell that? Um and and in periods of peak demand, you sell it for a high price.
And when the the demand drops, you elastically change your pricing schedule to, you know, respond to the market. And so when you buy a computer like I've got an M1 laptop here, I've never used it really at its maximum capacity, but so I've never been able to really amortize the cost of the outlay other than some efficiency gains in work and ergonomic comfort of of working. But if I could, you know, buy a $1,000 Linux box and sell every other every part of it into a an edge sort of whether it's edge storage like as a elastic, you know, mesh CDN or whether it's, you know, yeah, cing for that, whether it's um doing selling cycles in a distributed supercomput that's crunching data sets for scientific things. um whether it's you know routing bandwidth to to create the whole street having a Wi-Fi network that you can walk in and just tap into if your cellular network drops down um and you pay pay off per from your neighbor like all these things are available with Bitcoin offering payments down into such a small level and that's what that that's kind of the reality that's on the other side of saturation of of the adoption of the technology in my view And it it guides in the ideation of product and and stuff like that. What what can be created is anything that allows you to to sell surplus energy, surplus resource you have into a market where it can be brought down to even Craig's recent article saying sub Satoshi payment channels.
>> Yeah. So if you could you just need to settle it at 100 millionth of a $12 coin.
>> Yeah. And it always and service providers always devolve into identity issues too because you know if you offer up a service you might have legitimate customers but then you might have somebody that's trying to do something illicit. And it always, you know, all of these solutions have to have a real strong um accountability foundation. Um, which is actually what makes BSV so attractive, but it also, you know, it it puts a lot of onus on the business owner to take care of that kind of thing. So there's actually a lot of work to be done with identity and >> all kinds of other solutions, >> you know.
>> Yeah.
And that's the is like you know in the last say two years with the proper SPV plumbing um and we've we've kind of got proper SPV plumbing but it totally changes the architecture of an application and the user experience and it's still very you know like it's it's improved a lot but it's very clunky in its own right into what people are familiar with and um It involves someone willing to embrace that friction to enter into a new paradigm and they don't often realize that you know what what a peer-to-peer paradigm offers them in terms of data ownership and privacy and such. So you you've got to sell the value proposition but still reduce the friction, make things as familiar as they you know they're they're used to with a legacy experience. Um, >> yeah.
>> And a part of that is identity to to unify that and to allow you to exercise it and and reap the benefits of data ownership and such. And it's such a big fight because it's uh it's now at highest you know uh the highest levels of politic and government [laughter] and uh you know a lot of these issues are being hashed out right now with competing technology stacks and competing uh knowledge bases.
Um yeah, it's a very interesting time and uh I just hope that the scalable solution wins, you know. Um we're we're definitely entering into a new age. That's for sure.
the to me the the fundamental like contrast there really seems like every other solution that's been proposed um for say digital identity and um you know that that integration of say a CBDC with with your identity and with access to an IoT economy or um you know sort of cashless economy um they all seem to be an accountbased architecture.
Like I haven't seen anything that's talking about it being UTXO that I don't think I just don't think they're that they're fundamentally like divergent paradigms that people people assume that UTXO wasn't scalable. That's the whole point.
That's the whole error. Um and it's caused billions of dollars in uh development work that I think will ultimately just be scrapped. But but you know, you never know how people are going to react. you know, reality.
If you would have told me back in 2013, 2012, 2011, whatever this all, you know, we'd be sitting here talking about these kind of issues, you know, knowing what we what has gone on with BTC, BCH, and a million other uh a million other competing things, you know, I would have called you crazy. But here we are and u I never cease to be amazed at how things turning out. You know this is the long hard road for sure.
Um but it's also very similar to what I you know so I got into the internet very early too because my first internet connection was around 92 93 something like that. Um and uh then I started well started working professionally at the end of 94. So you know um I saw the rise and fall and rise of uh the dot era the startup that I worked at was one of the largest uh we they were the largest raise of the 90s of the dot era. They raised the most amount of money of anybody back then and it was wild because I was like the first technical hire at this thing and it started like 95 so we rode it all the way to well I I jumped off at 99 but it was you know anyway that's a story for another day.
>> Is it still alive or did it die?
>> Say that again.
>> Is it still alive or did it was it a casualty? It was the It was the worst invention of the 2000s.
It was really invented in the 90s, but it was the worst invention of the 2000s.
It was something called a QCAT, which was a it was a little barcode scanner that would take you to a website. You could scan any barcode in the supermarket and it would take you to a website.
And uh that was part of the technology.
That was the thing that everybody panned. They hated that part. But we had also what we had done was created early scandalink technology and the people I was working for patented all of that.
And so it was my first introduction to intellectual property, first introduction to a a wild startup. Um because it was also a TV show that ultimately ended up being syndicated in over 200 markets, 300 markets across the world. It was a TV show on computers and the internet. It was the world's first triple cast. So, it was uh simal cast on radio, TV, and the internet all at once.
We were one of Mark Kuban's early customers there at Audionet in Dallas.
Um anyway, I got Man, I got stories, dude. I got so many stories. It's funny.
Anyway, seen some wild things. Okay. And I also know that it took a long time for the internet to come back, right, to where we actually started to have useful apps.
You know, that that didn't start to happen until like 2008ish, you know, it was like web two, right? The whole uh the whole resurgence. Well, that took a long time. You know, if you consider like the beginning of the 90s is really the start of internet technology and people talking about it and then the commercialization around 9394 and then the dot boom bust recover you know 2008 that was that was the time that's almost 20 years. So here we are that actually was 18 years if you you count it like that. Maybe it's maybe call it 15, whatever. We're in the same timeline.
>> And that was that was, you know, massive amounts of money pissed up the wall >> and in blockchain, you know, like >> I I I can't think of a successful, you know, product that's been built on a blockchain.
There's some uh well it defi it depends on your definition of success but there are some professional blockchain enabled apps that I know are doing stuff like signatures legal signatures um things like that um that are just running silently in the background and you wouldn't know because >> yeah it's something that uh >> I mean if you can't identify the transaction then you don't know what it is.
>> Sure.
>> No voluntary >> 40 40TPS. So you know success is is something that achieve scale you know like there's businesses that are viable that that are running on the on the chain and that's success in its own right if your goals are just to run a viable business. But in terms of a flagship product or something that you can hold up and say this made use of the technology and and it's like smashing it. I I don't I'm not aware of any in >> Yeah. Um I mean even the thing that I'm building um the multiccast network actually works. I've proved it up to Okay. So I've got a 25 gigabit interface in this lab server that I'm using and I can get 16 million transactions per second on that server, one server. So, um I think at 400 gigabits I can probably hit 150 million transactions per second on one server. So, I would consider that successful. Uh cons, you know, because that means that I can I could scale the whole darn thing to a billion per second, no problem. All I need to do is stand up servers, order bandwidth, connect them. I already know the multiccast stuff works. I've got an automated platform that will automatically converge everything across to a fabric. Um, it even computes a a lambda 2 KPI from a what they call a plassium which is another thing that Craig wrote about. Um, digital cash is a network problem first or something something like that.
>> Well, I actually integrated that and it works. uh you can uh you tell it you say well if if this network segment's weekly connected it will it will flag it with an actual algebraic formula and it will it gives you a number and you know how to you know how to react to that number.
you stand up another link and you increase the level of connection, suddenly you've got more resilience, right? It's a way to map networks, >> patchiness, spottiness kind of thing.
>> It's it's spectral network analysis is what it is. And I think it was mostly developed for like brain networks and such, but you could actually apply it to a multiccast fabric, too. So, >> well, would you would you consider that a product on I'm try because like when Todd said he didn't he doesn't know of any product and then you described like the signature product like so you you know that's more like an app that users interface with right whereas what you're describing are users interfacing with >> okay so let's get to that part because that's some juicy details here okay so uh the core the core function of the multiccast network is to distrib contribute uh extended format transactions to Teran node miners. Now extended format transactions aren't used in the overlay ecosystem. So they're not used between users, but this network can take in a beef format transaction and send it to a completely separate set of subscribers on the multiccast network.
So now you've got these ingress endpoints all over the world. you're you're sending to your nearest one automatically and it just pops the transaction everywhere all at once. It's free for you to do, right? Well, you can do that both for minor transactions. So, those are settlement transactions say um or you can send a beef off and go to the overlay infrastructure. And actually, I just started up a discussion on the Babage Slack about this. Um, anybody who wants to participate in that, feel free to. Uh, we need a way to connect the overlays to the beef transaction format so that overlays can receive a subset. So, they're an active participant on the multiccast fabric, right? So, a consumer doesn't have to receive everything. You can subscribe, right? This is why it's layered. So the the top level group is for blocks.
Uh Coinbase transactions which actually have to accompany blocks and anchor transactions which are um think about uh a set of change transactions, right? Millions of them in a row and they've all got a parent.
Well, you send the parent transaction to the wide block channel so that all participants receive it. so that they've all got an anchor that they can use to verify all of those chain transactions, right? Think about an app, right? You don't want to be waiting on something that's missing, right? So, make sure those are all visible. Sub trees also have their own group channel for distribution um and the the transaction shards themselves. And so, you can do the exact same thing for beef also. So you have overlays on the network.
>> Can you explain what a what a beef transaction is? I know I've heard me mention it before.
>> Um, basically it it includes all of the ancestral information that you need to do a an SPV offline validation. Um, it basically just adds enough transaction detail that you can do extended validation on that transaction without having to fetch it from somewhere else. every every output that you're consuming, it carries its um Merkel path like or the the interior node values of the Merkel tree for it to get to its Merkel route of the block that contained that output. So then when you're checking the outputs, you can do a quick check like, oh, this one comes from blocks 832,000, this one comes from 745,000.
And if you've got 50 inputs that are being consumed, you'll have the beef payload will be 50 different um block different sort of like JSON strings of all the interior values. So you can recmp compute that Merkel route and check it against your local headers and as long as it wasn't spent in the last block, you've got confidence that that was, you know, um not a double spent transaction. There are ways like if you had 10 transactions that all came from the same block, you can kind of there's there's different versions of beef where it can compress them all in and sort of dduplicate those interior node values if it was from the left side of the tree or the right side of the tree as you're recomputing that. So that's like a low bandwidth um SPV thing that I think Craig published an article about as well to and there's one of the BSAS is a V2 or V3 or something like that of beef to >> make it around >> and remember everything outside the miners the overlays the apps the users need to be using SPV. So the the reason beef includes all of that stuff is to make SPV work. And when you do that, you can um basically to recreate the pre-image that needs to be signed of the um the previous transaction when you're hash signing the hash of the previous transaction. Um you need certain values that the miners used to have to look up to recreate. So in the extended format you've got you provide these additional values that are required to reconstruct the pre-image and that's an extended format transaction. So the the miners receive it and if you've given them that extra stuff, it saves them having to do a lookup and so they will kind of much muchly greatly appreciate an extended format transaction being handed to them because it stops them having to go into their UTXO set, find the um pre or whatever it is >> for the the acronym. It's an acronym, right? So >> extended format or BE background evaluated extended format.
>> Okay.
>> Yeah. Yeah. An extended format is just a a format of transactions that Teranode uses. Um it's got the the the locking script of the inputs and the the output value. They teranode needs those two things and if you can avoid a lookup.
>> So if you can avoid having to go look up that that's how you get high throughput is is avoiding these little things. It's these little optimizations um at the various layers. So >> originally didn't even value in I think in the first like sash algorithm there wasn't even value in there.
>> Yeah it's well BRC12 was never you know it's just the transaction it's just the script it's just the the you know it doesn't include all the information from the input so you have to do a secondary lookup it's just because it was the old way to do it in Bitcoin D another super basic question. So you said BRC 12. I've heard, you know, BRC 21. I've heard of 47 and I've heard of >> Yeah.
>> What are all of these like >> BRC? You know, >> the way >> the way the internet works is they had something called a request for comment or RFC's.
>> BRC. Okay. Got it. Yeah. Yeah.
>> RFC's. And this is a Bitcoin request for comment. So, it's just a way for engineers to to put out a memo stating, look, this is how you can build a certain thing and I'm soliciting comments on how to make it better or you know, whatever.
>> Thread like a basically like a thread like a topic, a themed topic.
>> Um, but it's from a technical level, right? So, it's engineering related stuff. So, if you wanted to define a a protocol, right? Um, for instance, BRC12 is a definition of how to construct a transaction, send it over the wire to a Bitcoin node, an old old school Bitcoin node, right?
>> Yeah.
>> Um, that that's all it is. And and they all they can be different, you know, they can they could describe different things. So, >> Got it. Got it. So, there's not there's nothing is deeper into that like there's not a 12 A B C. It's just 12 13 14 15 and just keeps climbing and climbing kind of thing.
>> Yeah. and they're meant to supersede each other. So if you do some extensive um work on whatever a BRC number and it's got a substantial amount of change, we'll just publish a new one, you know, and refer to the refer to the new one rather than the old one. Um it's just a way to define engineering standards so that everyone is on the same page. when the when the BS when it was the Bitcoin association back then but they had the technical standards committee and the idea was to formalize these things um as technical standards but it it became sort of a maybe a little bit premature and bureaucratic process and it wasn't that that optimal in the way that it converged on a um a reality because it said you know for the first um standardized Merkel proof it kind of A couple of people who were on the committee thought about it. They invited some industry involvement if anyone wanted to suggest it, but it had minimal involvement. And then they said, "Okay, here's the standard. We're going to roll this out." And a bunch of people were like, "That doesn't serve me. I'm I'm now bitter about that." And that happened a couple of times and it basically just killed the the process and said, "This isn't this isn't working." And then the tire really proposed the BRC method and and seeded it with maybe 70 or 80 BRC's that he he put in there. And um that allowed people to debate, you know, the merits of particular ones, extend them and and and suggest, oh, let's do it this way instead. So certain ones became >> Is it like a proposal kind of system or >> It is a proposal. Yeah, that's all it is. It's just a prop and then >> most of the time someone's writing some code that uses that proposal, you know, and puts that out also. That's what I'm doing is I all I've done is taken a bunch of um you know, I've defined the different multiccast frame formats or the wire format that you need to put on the network to for the proxy to use it to send it along to the to the listener and to the consumer.
So it's just defining that and and kind of standardizing on it.
>> Got it.
>> Yeah. But this it's really led to a whole lot of the um the growth in that you know the metanet based architecture is having having those specs and especially as everyone's now um or most people are are building with a pair program or with an LLM or something um being able to just shove you know get give them context around these ways of doing things because formally everyone would do something bespoke oh yeah I've got this derivation scheme or I've got this you know way of way of um you know broadcasting a transaction and if you just everyone's shoving you know we're doing identity certificates here's BC52 um you you all know you're kind of working around something BRC 100 like it's so good now because I'll sometimes use the metanet wallet I sometimes use the BSB desktop wallet I've got my own BRC 100 wallet I can Bridget can run it against her code um John can run it against his his wallet and It knows that create action does does that. It knows that internalize action like internalize a transaction does that. Um so it's allowing people to build things that are going to be interoperable and avoid a money button type situation or avoid a um you know custom derivation scheme.
You know like TW there's some talk of Twitch coming back up now. I was like, man, I don't even know how I'd um log into my money button account to, you know, get that key to >> see the thing is that you need to do it SPV now.
>> So, like, you know, >> Yeah.
>> Is it is it just that they're going to revive the old, you know, ways of >> I don't know. I don't think the thing really worked super well before. I mean it worked but the idea of the payments it was too much friction. See they they need to reduce friction. They need to reduce cost to where you don't even have to think about it.
they're anchoring it to a um to an accountbased model in there like you you there is you know you do need to have an alias to be addressable but it should be sort of managed in an overlay and it should be against your peers and in a sense um like I can be in someone's contact book as Todd and I'm in their in their PKI or whatever, they know that Todd is me and someone else can have a different at Todd and you don't have the conflicts of an alias because it's just scoped to your own PKI, your private key infrastructure and your contacts and the peers that are relevant to you. So you've got a a simple alias rather than a key that's mapped to their, you know, a type 42 derivation scheme and a key rotation, you know, schedule. But um, you know, you can ergonomically just address them as as an alias, but it's going to be a different key every time. So I don't think Twitch was originally architected in anything like that. It was and it was using an op return payload rather than anything that's a spendable token and um >> it was pretty like it was polished but it was still friction and it was still it didn't make the most use leverage the technology in the best way in my opinion >> and if the data is not federated meaning I can't like move it or portable I guess is the more accurate term uh if I can't move it to another platform that kind of sucks. Right. That's a huge pain point now because you know I've got all this content on X but what if there's a new you know what if I want to move it over to to >> that's the whole promise I guess with like the metaanet isn't it? It should be >> should be because we own our outputs, right? That's what SPV is all about, managing and owning your own output.
>> I saw that too that they just in the last basically this last week they're popping up and I figured they they were coming back because of the new infrastructure and stuff, but like it would have been awesome if they really were coming back to run it in a really Bitcoin native way as opposed to just coming back for the sake of coming back. Bitcoin. It wasn't just a It wasn't a peerto-peer native way. It was >> okay. Yeah.
>> It was like It was like BSV 1.0 kind of way, you know? We had the scalable black scalable blockchain, right? So, throw everything on it, >> right?
>> You know, but use it like the old Bitcoin, right? Well, we can do that. We have SPV now. We've got all the plumbing for that, right? We've got the >> So maybe they >> maybe they just need to consult with uh with you guys and >> they they just need to put it in BSV browser and start building it the right way and then build some killer features that people actually want to use. You know, >> build silo and and an account. You build it as a client that allows them to traverse an individual's own data and in a in an overlay that is that, you know, protocol. It's the Twitch, you know, a Twitch protocol. So, it's great. You know, you've got the a slick skin that allows you to navigate all your content that you've you've submitted to an overlay that they they manage, but you can at any point sort of take your, you know, 0.5% of the contributions to that overlay and send them somewhere else.
They still have like a carbon copy of your content on their overlay. Um, but now you've you've taken your data and said like I I want to move it to I want to also publish it in this overlay and I want to use this Twitch browser which a Twitch sort of client like how IRC used to work. You would just have you'd have Adom or you'd have a different sort of client that gave you different features um to publishing messages to your peers.
That's how like that's the sort of purest architecture that I'm I'm more like interested in and really hoping to see someone produce. Um >> yeah, but then also there's private privacy things that they could do. You know, there's all kinds of things that they could they could roll up into that thing. and you know >> time will tell >> from Babbage crew or like hope hopefully they've you know embraced that and and it is a 2.0 or something like that if it's just resurrection. But um >> yeah, >> you know, >> I'm sure people will jump on. There's some real There's some real loyalists in there that will >> I know. Don't think we're bagging on Twitch. We're not. It's >> just >> I used to like all those funny memes, you know, that the dude would What was his name? The guy that was >> Yeah, but what was his the guy's his real name? The He was at CoinGeek a couple of times, the the conference, right? Petty >> Josh Josh Petty. Josh Penny. Yeah, he was funny.
>> I don't know.
>> Anyway, you know, we could get more of that energy back. That'd be awesome.
>> But with some good apps, really good stuff. But, you know, that's it's tricky with those ones like um if they're going out with their own client like um you know, Tree Chat's awesome, but I just forget to go back back on there. Like I don't I don't really want to get sucked into another like app. I'd rather have this efficiency of everything's coming through one app and I discover like >> check everything that's relevant to me that's on the metaanet. This is why portable data is so important because you, you know, if you're building a corpus of work, you don't want to have to leave it behind and go redo everything every time, you know, or every 10 years or whatever, you know, it should be portable. We should be thinking ahead.
>> Yeah. And I I think everyone should be, you know, instead of paying your $10 a month subscription to some crappy SAS, everyone should be running a $5 VPS and that's their story a little bit more, Todd, about >> Yeah. Just explain a little bit more because this is kind of what what um John was talking about, right? Well, so like I've got um you know at the moment I pay $2 a month or something to Google Cloud to get my 100 gigabytes. Um but then I also you know let's say Dropbox or a couple of other things like that.
Maybe there's some, you know, storage that you want there, but I'm still using their infrastructure. So, like if I have a a VPS, a virtual private server, I can get one for $5 a month that's got 20 gig on it and, you know, a crappy little computer on it or se segment of a computer, but I could be using that as my cold storage. I I could get a increased storage plan for $10 a month or something. I still have a crappy computer on there, but I've got now, you know, 80 gig or something. And all my content that I create on the metaanet is kind of mirrored on my VPS. And you know >> it should be >> it should be mirrored on multiple storage providers that compete for your business. And actually >> um I would like to see >> uh finding a floor a pricing floor for everything. You know, so much of this world is focused on pricing higher and higher, more spend, right? I'd like to see the opposite. I'd like to see things getting so ubiquitously cheap that >> storage providers are falling all over themselves to get you to store their data and all you got to do is spend a few satoshi's to do it. Um, >> and you pick three of them and they're done. And the thing is is that this, you know, have to bring it back to the multiccast network, but I'd rather send one packet and hit all of them at once without having to without having to even think about it. And I I feel like a lot of these application distribution issues and resiliency issues and um just the architecture of the way that they work, I think that they could benefit from something like this because of coordination, right? Unpack it everywhere.
>> Yeah. And dduplication.
>> Exactly. Exactly. And actually that's kind of an inbuilt feature of this multiccast network that it uh it can dduplicate TX ids.
>> So if you had u so it can actually enforce the first scene rule at the network level.
>> Um whether or not that's a good thing we can debate that because there's a latency right across the planet. If you injected the same transaction at both ends, you could, you know, if one was a few milliseconds faster than the other, well then certain terror nodes that are closer to that side might get it faster than this one. And then you might still have the same problems that Teranode has to work around with competing transactions and whoever wins the block's going to win. And you know, it's got all that. But you you could do it.
But if you if you you drop on there is a doc x who's going to double spend your you know your changes to save changes to a document >> and this is why small outputs matter mostly because they they become economically uninteresting >> they are not worth trying to fight for them right um >> yeah um sales force or something like that and they want to do you know some kind of intricate embezzlement And if you send that purchase order quicker than that one, you can frontr run it and you know get access to a contract before like that's that's where the double spends can have a significant economic upside in a giant world of like ERP or EDI or something and it's like if I can double spend this thing and you know get an approval and blah like but that's up to the business to manage that side of things with you know data access management or data event management which can all be done on there and you just have a you know 10 minute settlement window or something before a particular and it's that um when Craig talks about it being ep epsilon bounded SPV you know just cap cap theorem mitigation through bounded epsilon you know thresholds that you're saying >> of a small value casual payment you know you can have almost an absolute guarantee and from higher value things or high higher value sort of consistency requirements or availability requirements under cap theorem. You just go for like a less optimistic eventual consistency and reconcile it at a 10-minute interval or a 60-minute interval or something.
>> Yeah.
>> Don't allow down >> mutations on a on a >> unconfirmed spend yet. There's a simple way to treat it also and it's that was a topic of conversation developing teranote is if you've got a double spend on the network there's only two things that it is. It's either fraud or it's a bug. So just drop both of them.
>> Yeah.
>> Put them in the sandbox for a while and don't refuse to deal with them as a as a policy. You know I mean let them submit it later but just refuse to deal with it. It's not an immediate blacklist that IP. It's just like, you know, time out, you know, send it send it the next block. You you let you diagnose why it's not working. Either >> improve your fraud or um or sort your bug out because it's Yeah.
>> So, I'm pretty sure that is a setting in terote you can you can enforce that.
>> I remember that being declared. Yeah.
>> I have a question for you. Uh Jeff, is there anything about Bitcoin's original design that you think could be better or tweaked that would help the design implementation of Teranode today? Like you guys are just talking about like these 10-minute block times and stuff like that. You know, there's blockchain millisecond block times and second block times and stuff. So was curious, you know, if you had any questions or doubts around the original design of of Bitcoin.
>> No, because Craig has explained it incredibly well over the years and I know that a lot of the source material that he wrote is not available anymore.
a lot of the talks aren't available and and people maybe newcomers in the space over the last couple years may not be aware of all of the technical arguments but I was pretty well convinced over 10 minute box hard >> chap you know summary of and we take us back to where >> uh I'm also not the one to >> you know what that's something that like Brendan Lee or, you know, Connor, you know, one of those guys could really uh could really articulate it well. But, you know, to to be honest, man, it just works. Um, when you when you look at the scaling path that we're on, once you fix the network, >> mind you, which is what I'm working on.
I believe so strongly in this that it really needs to be done in order to >> you say that once you fix the network because like there was a question I wanted to ask early on but I think I guess where right now it might be relevant which is like what you know what is the difference between the ter node now versus the BSV node from before is that what you mean by fixing the network? Well, that's part of it because we had to get rid of the legacy RP the not the RPC interface, but the the legacy peer-to-peer I think is it's called it's actually called legacy in Teranode. Now, that's the way to transmit blocks that contain all the transactions in them and then a node verifies everything after the fact, right? Well, in Teranode, we flip that upside down. We validate transactions continuously. We wrap them up into sub trees every second and we announce those sub trees to the network. Other miners pull them. That way they have groups of pre validated work to quickly attach to announced block templates. So in a a template a block template gets announced to the network. It's got all of the pieces of the 80 byt header. So it's got the Merkel root hash. It's got all of that stuff. And the Merkel root hash actually rolls up with all of the subtree identifiers which are each hashes.
They're they're each submerkel root hashes basically. And all of them add up into the same cryptographic tree basically. Um but what that allows a node to do is to quickly verify and move on to the next proof of work. right now.
>> Sorry. As opposed to as opposed to what?
I didn't >> As opposed to doing all the work after they received the block.
>> After they received it. So, they're doing the as it go attack vector. It's an attack vector.
>> Yeah.
>> The old proof of concept I feel like was done intentionally because, >> okay, think about this, >> right? The way that they the way that the hash wars work right now, somebody with an overwhelming amount of hash just keeps producing blocks over and over and over and before the lower the before someone with the minority hash can even can even validate them, right? Or can even uh can even do a new work, right?
That it's an attack vector. And the BTC guys talk about this all the time. Well, With Teranode, you're doing the work continuously. You have to do the work.
You have to verify the sub trees. You have to download them. You have to check them.
>> You have to you you have to transmit them to others. Other people can see your work. other people, other terror nodes, other miners can see your work as it's happening because you're you're you're transmitting this information continuously every second in the subree structure, right? So you can make a and this is where Tai's arguments are brilliant because you need miners are going to have to start doing uh they're going to have to start looking at that actual proof of work to see what miners are expending resources and which ones they should be paying attention to and which ones they shouldn't. And if a minor that's got a ton of hash power but very little work that they're doing, right? They're not they're not actually uh linking all of these transactions.
They're not validating them, doing the hashes, building them into the trees, sharing them with others.
>> Pure form of proof of work in a way, right?
>> Yes. And you can start to do an objective evaluation of the people that you are building on their work. Right.
So if this so if this adversarial miner is is advertising these block templates which are sparse which are lean which don't have all of the sub trees are not doing all the work what are they actually accomplishing well not a whole lot if the other terodess are paying attention right so just by changing the peer-to-peer format I think I mean it's not it's not perfect I think you could still game it you know if you had a bunch of hardware whatever but it it's a different paradigm because you're suddenly you're shifting the proof of work to not just after the block. It's not just block oriented, it's transactionoriented. It's it's everything in between, right? Um you you know, a lot of this stuff isn't built into Teranode yet, you know, evaluating other sub trees. some of it is, you know, they check it for technical accuracy or whatever, but um I just think there's a lot that can be done with those structures. And I think there's a lot that hasn't been talked about with sub trees. I think that there's a lot of magic there that has not been developed yet. Um especially for like users, for transaction specialization, so specialization of miners, >> um things like that. Um the network format that I have includes a subtree ID. So you can attach a subree ID to a transaction. Um and then it has the ability to filter based on those subree IDs at the edges. And I created kind of this rudimentary subree group announcement protocol to where you announce group membership of new sub trees and by that you can automate certain infrastructure um basically on the consumer side so that you're only delivering a subset of the traffic. Now, this is the part that wasn't explained really well in the multiccast and multiccast blog, and it's the part that I've kind of had I've struggled with because I think a lot of those solutions are still being kept tight lipped right now.
>> I think we have just the barest details of what sub trees are. They're expressed in the go subtree library now and in tyrannode but I think that they can probably do a lot more. So >> that was when I wrote up that Merkel trees course on the BSP Academy that I actually didn't didn't formalize it as a sub tree, but I was saying that in the future I'd imagine that the a node is there and say transactions A through to F uh hit the network. Um you might receive them. You're a node operator. Um Anil, you're a node operator, Jeff. And one might have it as like a c bf.
The other might have it as a e c b e f.
Um depending upon the the different way that you've received them. So your merkel trees are going to all have a different route. So I'm there as an operator >> and I've got an you've streamed your sub tree at me. So I've said okay that that contains transactions A to F. They're all valid but yours has a different ordering. So you've got a different um Merkel route and Jeff yours has a different ordering but Anil you um then the blocks produced by you and you've also got a transaction G. So when that comes into me with with the block it's like here's you've got subree one and you've got subree two which also has transaction G in it. So I need to then just pull to get the missing ones that have occurred in the last interval that weren't on that first subree. But I need to have sort of ced your Merkel tree the merkel tree that you're built I'm building for you and the same ones that I'm building for Jeff and all the other peers that are worth connecting to. So I've got sort of building building maybe six in parallel from my highest confidence nodes that I subscribe to their trees and they've got a different ordering but I 99% of the transactions are in there and then at the end I just pull for the missing um ones that have came in before that last subree announcement if the block occurs you know in a in you know discrepancy between that is that sort of is that occurring are nodes building the other subre Yeah, Teranode um orders sub trees from its own vantage point and every teranode has its own sub trees and they have to share all of those sub trees with all the other terodess. So actually the multiccast network helps here too because a minor should only have to send that subree up to the network once delivered to everybody rather than deliver to all of them. All of them have to I mean subree data is not very big at a million per second it's like 32 megabytes per second but it grows you know 320 megabytes per second when you get to a billion right or is that right >> yeah well what [clears throat] >> you're right you're right it's a lot more than that yeah um uh so it becomes a data problem on its own And actually it becomes even more of that. And part of my work on Teranode was um figuring out low latency storage shared storage systems for the cluster to use because those sub trees get stored as files. Um when they come in it's just a it's just binary data and you need to be able to share that with all the services because all services need to have access to all sub trees continuously. So a lot of the work that I did was with luster file system um and bribing that low latency uh shared storage access to it. Um also in the later revisions of our work I was working with uh tuning NFS to do the same thing but I think as it scales up it needs a really robust storage system. So uh that's its own that's its own distributed problem right storage.
Aerospike that's aerospike was the UTXO for the UPS and then this is for the subree storage >> there isn't it event stream or is it something that's >> yeah there's the there's capka for the message bus so >> uh that's how terror node coordinates all of its operations between all the different services shared storage is for the quorum data subree storage uh transactions which won't fit in aerospike have to go there because aerospike actually has a limit on the record size and so they had to work around that issue with it. Um aerospike handles the UTXO store, Postgress handles the actual blockchain database.
So just the blocks and all of that all of that data. So it's actually a bunch of different systems that have to tie in together.
>> Do you know Derek Moore?
Yes, that sounds very familiar.
>> He had a system PAGA and back in the day like when when Shatters was doing um Teranote, he said they're not going to be able to scale it with CFKA. They need Prove need to come and talk to me. So, I know >> and they're actually not using CFKA now.
They're using something called Red Panda, which is a C implementation of CFKA. So, CFKA was built on Apache, Java, that whole ecosystem. and somebody came in and said,"You know what? This doesn't work at all. Let's redo this and see." And they launched it as a commercial open-source community and commercial offering called Red Panda.
So, it's way more scalable. But you're right. You're right. It only scales until a point. So, that you know, there's always going to be a need to look at the next layer, you know.
Um, >> and there's so much to optimize, too.
like we haven't even got into like optimizing script you know and uh Dr. right has also p uh published uh hardware methods of it wasn't floating uh it wasn't gate arrays it was um it I think it was AS6 uh to speed up the validation process greatly so you put all the op codes in silicon basically and so you're so every transaction takes nanoseconds to to validate now so and it's a card you and pop into servers. I mean, these aren't built yet, but the the the foundation is already there. He wrote about it. He he posted Verarilog uh Verarilog files with it like anybody could go and build one. Um I mean, as soon as we have that, you start to optimize that part, you take a spike and you you take that magic that they're doing with NVMe storage. See what they're doing is they're exploiting key value store built into the storage networking industry association uh tooling that they've come up with. It's a direct interface to the NVME device.
So super fast key value store >> because there's a CPU on the NVMEs, isn't there? So they can actually have like a little lookup against the M map or something of the file there. it it's basically a it's basically a hash index lookup on the hardware itself. And so Aerospike is taking full advantage of that and getting you as close to the hardware as you can. But >> yeah, >> there's nothing stopping anyone from building a purpose-built UTXO store that does kind of the same thing. Um when we were talking to Aerospikes people, they said that they had that 280 million transaction or query per second. um commercial customer that they had. It was running on 1,200 servers and these were like not high-spec servers. These are like commodity boxes, but 280 million per second like you can >> Yeah.
>> you know, it scales.
>> Yeah, >> it scales, right? And if you can do it even more streamlined than that, because Aerospike does a lot of things that Teranode doesn't even need, you know?
So, strip all of that stuff out. Make it handle just UTXO records. Just do the operations. Just focus on efficiently implementing every operation that needs to happen on the back end. Every single piece of this is designed to be scaled.
Every piece, the whole design. I' I've never run into anything where I think, ah, yeah, that runs into a wall right there. You know, never been able. And I haven't seen it with this multiccast solution either. you know, as I was a little apprehensive about when I first started it, like, okay, is this thing actually going to work? But I've got it in the lab and it actually works. I mean, it it I've the system that I've put together has um reliability at every every tier. So, at the ingress level, it uses BGP anycast, multiple ingress proxy points with fail blast blast zones for BGP basically. um so that it fails back graciously at the regional, the local, the regional, the geo level. Um >> on the consumer delivery end, the way that I've designed it is two tunnels, either wire guard or GRE tunnels, either IPv4 or IPv6.
I do an A and a B tunnel. The feed only ever comes on one at at a time, but I use portable IP addresses between them. And I use portable I use BGP at every layer so that if one component blows up, it automatically routes the feed to the other one and the retransmission system automatically replaces any missed packets. So there's no gaps. So it's 100% reliable delivery from end to end even if the entire data center melts down, goes on fire. So, it's designed to be completely resilient at every at every corner. It also um well, I guess I'll wait for Todd to get back to get into more juicy details, but um it needs to be reliable. That's, you know, it's >> it's crazy you were talking about like um what what Craig mentioned or not mentioned but like wrote about that allows us to optimize just the physical uh devices even further than AS6. And there's just thinking about all these >> pockets and and um areas of opportunity to optimize it. We're already at one million and like you know the path to getting to 1 billion is seems like >> it's not Yeah, >> it's not a big technical problem. I mean from Teranode there's a lot of things to to solve. I'm not going to sugarcoat over you know the backend part of it because there's a lot to it. Uh the block assembly service needs to be made more resilient for instance or moreable I guess. But a lot of the pieces are there to where if we just start applying these optimizations at every level bit by bit and then we get the throughput up high to actually exercise the system to to find all of these edge cases and to to get there 1 billion per second totally doable. I mean I could I could build that network today and I could demonstrate it to you. It it does work.
It's not fanciful from the from the transaction distribution point. I 100 gu 100% guarantee you it works today. Um it >> if I had enough hardware I could do it for you right now. And in fact if anyone wants to help me get there >> Yeah.
>> help me.
>> What's the what's the uh the the ask the plug man? Go for it. Go for it.
[laughter] >> You know what? I I just want to see that. I just want to see us actually scaling this thing. You know, it's scale or die. And I'm all about the technology. I'm all about the engineering path. Um I just want to see us succeed because this is the most important mission on the planet. I feel like um we need we need full accountability at every at every layer of the human existence. I should say, you know, that's a little dramatic, but you know, I' I'd settle for just, you know, scalable electronic cash.
>> So, we had Bridget on the other week and she was talking about how where Bitcoin sits in the OSI stack and she said, "Oh, you know, at the moment, it only sits in the application lay layer, but >> not true.
>> Not true. It can actually insert itself between every single layer." Well, that's you're talking about with with V4 and the way that it currently is. It really only is at the application layer with IPv4. Once you embrace six it um >> I was saying I think I think it sits at all the layers but you know with a transport say it really only sits there with um it would need a different wire format or something like that to embed you know um sort of economic routing to the the you know the header of UDP or TCP or like >> well >> you know having kind of extra routing table there for Right.
>> Uh it doesn't even need that. I mean, think about Okay, so layer three is the addressing and the I it's the IP layer basically. OSI layer really doesn't map to what we're doing today first off because it was written in a different time from a different engineering viewpoint etc. But you know layer three >> what is OSI layer is that?
>> Uh it's the open standards institute layer. Um it was a way that they came up with I think in the 70s to start to do engineering um and it was sponsored by the ANIE people. So a completely different track from the ITLE E family, the IETF. I'm not the I'm not a historian on all of that stuff. I just know that they had a different path through from different engineering >> like part of this uh stack right this internet stack where it's like physical layer then there's this uh network layer then there's the correct >> port layer and then the application layer >> it's the it's the it's the physical layer the the link layer which is Ethernet basically right >> the network layer IP on top of that and then transport which would be TC TCP, UDP, and then applications and stuff on top of that. But I, you know, we could do payment. There's nothing stopping anyone from designing a protocol for Ethernet, an extension that can do payments across a across an IP link. And there's nothing stopping anyone from doing the same thing with internet also. And if you and if you look at Bitcoin certified addresses like the some of the work that's being done, um that's player three, right? Um Bitcoin could actually be used as its own transport protocol if you really want to get down to it. You know, we could embed IP packets inside Bitcoin transactions. What you know, you could do all kinds of things. So that's what that's what Bridget was saying was that like Bitcoin right now only exists at the application layer, the very top layer, but may not to say that it cannot be further developed into these deeper layers though.
>> Yeah, it it's meant to transcend all of the layers. It doesn't mean that all the work has been done, but I would argue that we're already doing layer two or sorry layer three work.
>> Layer three.
>> Yeah. Um in addition to the to the application layer because u CGA plus is IPv6, right? part of the addressing standard >> semantering with that like sending direct seamanto cells. So they're the wire format, they're the storage format, they're the compute format. So it's it's giving some very interesting possibilities in the way that like for example when the cell is in in storage um they're 1 kilobyte cells. So you bring up a database page 4 kilobyt you just grab that off you know the 1 kilob chunk and you you don't have to do write amplification where you would grab off you know the 4 kilobytes then take it up just grab that 100 kilob 100 bytes there and you know to make the edit and leave the rest of it in in RAM while you're doing such things you bring it from storage you send it out the out the network port you don't even have to cross the motherboard um you don't have to serialize anything.
So like that's just experimentation and seeing what happens there. But it's there's once you embrace like BCA that's derived from the keys and if you have some you know there's in there's a previous hash of the you know the state of the cell. So you have almost like a monotonicity there and a you know a sequence number um or a sequencing. So you could get you know retransmission of missed intermediate states. Um, so that's like that that's, you know, it's using UDP or it's using it as a custom wire wire format just on a little mesh on a a LAN that I control. But that that is touching that layer in the in the stack.
And, you know, if there was two compatible routers on either end, they would be able to go great, this, you know, is still intact at the other side and decode it.
>> Yeah. I do some um some optimizations with slices and and mem copy to keep the the proxy path zero copy as much as possible. Um and also I shunt it to user space networking and the commercial solution. So the the open source proxy is just a kernel kernelbased solution.
It tops out at maybe >> half a million transactions per second on a six core PC, right? Which is actually pretty good. like you know it would work for our needs but the the commercial one on the same hardware will do 16 million um using a combination of user space networking uh address family XTP and uh a bundling frame format. Um most BSV transactions are small. They're like 216 bytes on average for pay to public key hash transactions. And so you can stuff a bunch of them inside a 1500 or 1440 NU packet. Um so you actually get more throughput that way. Um the the multiccast networks designed to accept single packets in bundle as many as they can over the over the links and then single packets out. Um uniccast and uniccast at the edges multiccast in the middle.
>> Yeah, that makes sense. And >> you know even if you you're talking about Ethernet there but if you get access to a laser or something then the frames can be bigger and you know then like there's that there is capacity for 4.3 GB jumbo grams.
>> Uh there are but actually that's a problem. Um so they did experiments with MTU sizes. You know, we have jumbo packets which is like 9216 uh bytes or whatever that can go on the wire. And they did experiments about 20 years ago with super jumbo packets which were 64k. And you know what they found?
They found that there were problems with CRC32 that it actually tops out uh because it can't it it doesn't have enough uh uh uh collision space and some other technical factors like it's not robust enough and we use CRC32 hardware error correction for everything. So they just decided well we've got to you know you've got to overcome this and it's a it's a hardware problem you know chip manufacturers need new hashing algorithms inside the you know new tech algorithms etc. So it actually becomes more problem than it was worth just sticking with jumbo packet sizes. Is that saying like essentially like the boundaries of the pixel of the frame or something like that? You you get bleed there or something like >> it's got to compute a check sum of of every single packet, right? And when the packets get too big, I'm gonna I butcher this explanation.
There's problems with CRC32 with super jumbo packets. Let me just leave it at that because I I'm going to butcher it's too late at night to >> try and uh to try and explain that. But there's papers that from 2006 or whatever when researchers did some of that. I'm sure there's been other work done, but that's kind of why they're sticking with 1500 MTU because most network devices that's kind of the base that everyone's always used. A lot of people hate it. It's it's it's caused all kinds of workarounds, fragmentation protocols, and I've got a fragmentation protocol for the multiccast stuff because if you have a 10 megabyte BSP transaction, you need to split it over multiple packets. So, um so there's a I can't remember which BRC number it is, but there's a a fragmentation packet format, but the thing is the user shouldn't have to do any of that. That's all on the network operator side, right?
So, I'm aiming for EF transactions in, beef transactions in distribute those separately.
No, no funny business otherwise. No new tricks.
>> Um, there may be something that we have to do to connect transactions coming into the network to overlays.
>> That's a discussion that I brought up on the Babage Slack recently. So, >> using maybe like version number or something there. Well, they use a transaction version number and there's four billion combinations, but that just gives you four billion address points to send those packets to. Is that enough?
Like, you know, and then who maintains the order? Like and if you pick transaction version 8 but somebody else picks the same one and you both send packets to the network and they're both hitting these overlays that both want. I that reasonable I don't know. I mean I see I'm I'm not in the overlay standards group [laughter] or whoever whoever that probably doesn't exist but >> not does >> it's not it's not my uh it just hasn't been my bag you know. Yeah. Um, so there's just there's things that we need to figure out there, but the the main use case delivering EF transactions to miners in a scalable fashion in a way that they can um if any one node on the multiccast network blows up, it automatically routes past it. It automatically repairs gaps. Um it's miners need reliability and I feel like by creating the right network structure we can benefit the whole ecosystem and all the miners at once. Like they don't they don't have to worry about network issues. You know that that's and this is the only way that it scales past a certain threshold.
you know, >> and the only way accessible to a business who comes in, they like they if if you get an enterprise to come in, they they don't want to be experiencing infrastructure issues on this. They they just want it to be like, you told me this scales to a million TPS and they want to just never face a hiccup on that side. Well, and I've built this network to where um uh a minor or a service provider could bring their IP space in and actually advertise their own ingress IP addresses on the same network to where they just ride the same endpoints and get distributed just like everything else, but it's using their domain name, their IP address, their everything. So, it's it's designed for the commercial path forward and like I said, it it survives interdommain ASM uh peering, you know, so it could it could grow past just one network. Um, yeah, I mean uh >> that's pretty well thought out.
>> I'm I'm trying, you know, I've got a I have experience in ISP uh design and and management and that's really helped here. And actually I worked at a startup few years back that had an aim to create um apps that worked all across the world and it would spin up resources based on need across the planet and it would just operate seamlessly on the same IP network and they had a goal of we were supporting like five or six different clouds all at once. Um, it was it was kind of cool and it actually gave me the experience I needed to tackle the organizational side of this, some of the BGP issues, some of the AnCast issues, um, because that's that's what we were doing way back when.
Um, so so, you know, it was actually right up my alley. I I'm >> really happy to take this thing on. I wish I had some help on it. Um, so far I've uh I've I've reached out to Dr. Wright a couple of times, but I got radio silence. And I'm guessing either that's because he can't talk to me.
Maybe because of, you know, I'm former BSV or something. BSVA. Um, I'm not I'm not really sure. Maybe he just doesn't have time for it. So, >> got too close to his his his secrets.
>> No, I mean, he put the information out there like he told how to build it, right? So, and he and he's the one who says all the time, just go out there and risk, build, risk, do the work, you know, put in the hours. And here I am.
Like I I heard that call, man. Like I said, was my dream job.
>> So, >> I I heard it loud and clear and that's what I'm doing. So, I hope that maybe maybe he could uh release a follow-up blog article that maybe answers some of those things about node specialization um how to use the subree ID to filter um on like are there different he mentioned hash algorithms that could create that subree ID like that could do different things right maybe what are those some of those things right Also, how does the NAR and DAR fit in with this? Because I could see a dedicated multiccast group subscription channel that's widely disseminated for network access rules >> and recovery.
>> Yeah, because that's really important too, right? So, there's there's just certain things that I there's there's pieces that weren't completely um discussed. I've figured out the hard ones like the retransmission stuff, the knackbased retransmission, um even rate limiting with that even solving the whole multiccast reflection problem. So if a let's say we have a minor consumer that's receiving transactions, they're also sending transactions up to the network. Well, they don't want to see those transactions reflected to each other.
So there's actually some really cool things that you can do with that composite hash key, the retransmission key that you can ddup that traffic back to their flow, back to their thing, so that they're not receiving an a mirror image of everything that they're sending out because by default that happens. So, I've had to >> I've had to solve some pretty interesting engineering problems and and some of the some of the BGP work where if an edge dies and their tunnel suddenly crashes, well, rerouting to the other one seamlessly, retransmission system fills in any gaps, new hardware gets spun up, second tunnel comes back in, same IP address, no reconfiguration needed on the consumer end. Some of that stuff is actually uh useful engineering that you know someone needed to figure out to make this work.
So um and it all scales up to uh data center handoff port connections. You know someday miners are going to be in big data centers and they're going to want to talk really fast to the network.
So this thing that I've built scales up to that. So 100 gigabit connections, terabitet connections, whatever.
>> Wow. Well, well done. Um, anyone listening, if you if you feel you can help in any way, you definitely get in touch with you, Jeff.
>> Yeah. Um, just go to 1bsv.net.
Uh, it's got a link to all of the open source code, GitHub repositories that I've developed everything on under. My company name is Light Web, Inc. Uh it's based out of Texas, USA. Um I'm just a oneman show. Uh you know, just wanting to keep up the scaling Teranode. You know, if if I wasn't on this, I'd still probably be working on Teranode and making it better and aiming for 5 million, 10 million transactions per second. But honestly, like I said, I think we need to solve the network first. So, >> uh it's it's going to be sooner or later. This is going to have to get done. So, might as well uh just >> get all the hard stuff out of the way while I have time.
>> Oh, well, well done. Um, >> anything you like want to cover that we haven't? Um, >> no, I'm just uh uh you know, I'm just here like everybody else trying to make Bitcoin better. And uh I I believe that it's uh truly one of the most inspiring missions on the planet >> and uh Yeah, >> I'm happy to do my part.
>> Yeah, cool. Well, I'll be, you know, I don't know how much I can [laughter] help with with the acronym soup that you've been tossing out for me. I understand a fair bit of it, but I I don't have that much to um to offer. Um but happy to be some eyes and peer review stuff and just, you know, give some feedback and at least send an act.
>> Thank you.
>> Yeah, >> thank you. Yeah, I'm I welcome any kind of input. I'm kind of working on this in a vacuum. You know, I'm the only one that I know of that is even in this realm of network engineering, systems engineering, um distributed systems. So, um I welcome anybody. Let's talk about it, have discussions. You know, I'm I'm easy to get along with and I and I love outside, you know, I love it when people tell me I'm doing something wrong because then I can go and fix it. you know, I I try real hard not to have a big head about things and just, you know, I'm pleasant to get along with. So, >> yeah, >> feel free to reach out anybody.
>> Cool. Well, hopefully that, you know, if you're listening and you've been excited and inspired by this, um yeah, feel free to to reach out or reach out to me and I'll I'll put you in touch and we can um you know, facilitate a a working group or you know, some collaboration here because I think yeah, what you're doing is very important. Um I think that there's there's not there's there's probably very few people in the world that in the world that can approach solving some of these problems whether it's the the legac having the the skill set in the legacy networking engineering and then also having the expertise of the um I guess the corpus that's been sort of published around the vision of of what can be achieved with the blockchain like a lot of the people who have seen blockchain so far that may have the networking expertise think that it's a dead end um think that it won't scale and we're showing that it will but we're doing So in a little um little veil of secrecy, not self not our own secrecy, but a um radio radio sort of blackout on our you know the communications of of the capabilities and and such. So um we I just want to get that the message out of what is available, what what can be done, where we're going with this. And if you have, you know, a problem that sub hundreds of a cent payments, you you would welcome it, welcome those in or you'd welcome data integrity solutions in the operation of billions of transactions a second. We have we're we're probably the only technology that I'm aware of on the planet that that can um that can provide a solution there. you know, not without some hard bridges to be crossed and some serious engineering challenges, but there's no um hard ceiling that anyone's that I'm aware of has identified saying no, this isn't going to work to to accept a planetary scale of transaction activity.
>> So, thanks thanks so much for coming and joining us today. And um anytime you've you made some more developments and you want to come on and it sounds like you there's more there's plenty more um you know we've only just scratched the surface of some of the what's what's happening.
>> We have we have like I said this is the thing that gets us to the thing and there's a lot more things that are tossing around in my head because once you have a network you can start to develop real services. So yep.
>> Amazing. Yeah. Neil, anything from your end, Dil?
>> Yeah, I've got a few tiny questions, but I'll leave it I'll leave it to after we wrap up here. I'll ask offline. Um, but I I love this deep dive that we did here with Taran. I I just, you know, I feel like we could do like a part two and and and go further and and I want to >> Yeah, >> happy to. actually when I get the terterod integration done um I just developed an engineering report shared it with a few of the developers um on how on what identifying the engineering gaps that Teranote has now to get us to this and it's actually not very much at all it's it's pretty simple um overall so I'd be happy to talk about it again maybe once it's uh a little more solidified maybe once the network is actually launched So >> cool. Look forward to it.
>> All right. So yeah, thanks everyone. Um please if you've enjoyed the conversation, like and subscribe and stay tuned for more content like this as we go and flesh out the the deep the technical details and and bring a level of conversation to um you know the serious engineering that's occurring on the the BSV network and the problems that we can solve for for anyone really.
So, thanks a lot.
>> Thanks for having me on.
>> Appreciate it.
>> Cheers.
Related Videos
27:20
Blockchain & Crypto
Multi Vendor Multisig w/ Seed Signer, Hodl Dee & QnA
BitcoinMagazine
985 views•2024-09-05
16:20
Blockchain & Crypto
Oasis Week in Review: Latest blog articles, workshops and more
OasisFoundation
135 views•2024-10-18
9:29
Blockchain & Crypto
Kaspa: How ZK Turn Blockchains Into Settlement Layers (Part II)
cxc
1K views•2025-12-19
14:40
Blockchain & Crypto
以言會友 EP13｜當比特幣屢破紀錄 區塊鏈技術能帶來什麼？
dotdotnews
293K views•2021-01-05
1:33
Blockchain & Crypto
Soroban Development: Ecosystem Growth, and the Rise of 70+ Smart Contract Projects
SorobanOfficial
1K views•2023-07-19
1:15:25
Blockchain & Crypto
Balaji Srinivasan I The Fiat Crisis | Pragma Tokyo 2023
ETHGlobal
37K views•2023-05-06
11:09
Blockchain & Crypto
$22 million NFT scammers arrested (insider evidence)
coffeezillaextras
806K views•2025-02-03
1:01:22
Blockchain & Crypto
SYMMETRICAL TRIANGLE HOLDS THE KEY TO NEXT MOVE" DON'T IGNORE
xrpfuturemillionaire
800 views•2026-03-15
Trending
13:38
Law
WOW! Judge TURNS THE TABLES on Trump in His OWN $10B LAWSUIT!!!
MeidasTouch
197K views•2026-07-23
13:25
Business
Playstation NO DISC/NO BUY Fight Is Over...
DavidJaffeGames
4K views•2026-07-23
22:40
Business
Steam and Xbox Just Dropped The Hammer On PlayStation
OhNoItsAlexx
9K views•2026-07-23
16:52
Anthropology
Americans Confused in Australia for 17 Minutes Straight
IWrocker
17K views•2026-07-23