LINUX Unplugged - 458: NVIDIA's New View

Episode Date: May 16, 2022

NVIDIA is open-sourcing their GPU drivers, but there are a few things you need to know. Plus, we get some exclusive insights into Tailscale from one of its co-founders. Special Guests: Avery Pennarun and Christian F.K. Schaller.

Winter may be coming to an end for us, but it's just beginning for the Mars Linux copter. You know, that thing's made 27 flights. Unbelievable. But recently, the little Linux copter on Mars has been struggling. Dust has settled on its solar panels, and it's having a hard time recharging its six lithium-ion batteries. You know, I just had this problem, Brent. I had to get up on the roof and clean off my solar panels.

Are they going to hire you to clean off the wings of this helicopter, too? Yeah. I don't know what the big deal is. Just get out there and dust it off. Yeah, just blow it off. Hello, friends, and welcome back to your weekly Linux talk show. My name is Chris. And my name is Brent. Hello, friends, and welcome back to your weekly Linux talk show. My name is Chris. And my name is Brent.

Hello, Brent. Wes is out on assignment this week. I think he's actually at the airport as we record right now. Well, coming up, we still have a heck of a show. NVIDIA has done it. They have kicked off the process of open sourcing their compute and GPU drivers. So let's chat about that. Let's chat about what you can expect.

Let's chat about what's happening first and break it down for you. And then later on in the show, one of the co-founders of Tailscale is going to join us. Let's see if we can get a sense of maybe what makes Tailscale so different by chatting with him. And then we'll round out the show with some boosts, some picks, and a lot more. So before we go any further, let's say time-appropriate greetings to our virtual lug. Hello, Mumble Room. Hello. Good evening, Chris. And hello, Wes. and a lot more. So before we go any further, let's say time-appropriate greetings to our virtual lug.

Hello, Mumble Room. Hello. Good evening, Chris, and hello, Wes. Hello, guys. Thank you for joining us today. I might sound a little different. I am remote on location. I am down in southern Oregon getting the very last bits of damage

from the Denver road trip fixed on Lady Jube. Her suspension took a beating on the way into Denver, and there's just one bit of repair needed still before we're ready for our summer road trip, because we are working on another JB road trip, and we'll have details on that soon. So I'm down here, and of course, it's right after my Starlink died on me. If you heard last week's episode, my Starlink is out. So I'm doing this an older school way, low bandwidth. So if I sound weird, that might be why. But and Brent, you're back at home now, all done traveling. I am. It feels funny that we kind of switched spots. You're traveling now and

I'm home. So it feels good for my end. This is funny for me, though, because I traveled over 400 miles. It was like nearly nine hours of driving yesterday. And I get down to Southern Oregon and I park and everything's where I've left it. It's the weirdest thing. It's like I'm in a different climate because it's much nicer down here. But all my stuff is in the same drawers. It's almost like when you close all the window blinds and stuff, it's like I really never left. I feel like I'm still in washington but

then i step outside of course i'm in a totally different different spot but i love it jubes did great on the drive down we hit some traffic but other than other than that she did so good and you know the cellular system i have now just so seamlessly switches between different networks that we remain connected the entire drive these days which is so awesome so we're you know streaming podcasts and stuff and listening to them on the way down. We towed the car. And so it's just the wife and I, the kids are at home because it's just a quick trip. And I'll be down here for a couple of episodes that I'm recording one Linux unplugged and one Coder radio, and then we'll be heading home. But it's

nice. It's nice to be on the road again. It's nice to be getting ready for the road trip that we're going to do later this summer. So we'll be sharing details on that road trip and office hours if you're curious. But let's talk about the big news. Let's just get right into this because I don't know if I ever thought this would happen. Every now and then, there have been these just ginormous Everest-sized mountains for the Linux desktop that we had to climb. And once we're on the other side, it's like life has changed for Linux users forever. This was definitely the case with Wi-Fi a long time ago. Automatic X configuration, I think, maybe could even get this treatment. There's been certain milestones that we have hit that have just changed life for

all future Linux users, whereas old timers look back and go, if you only knew what we had to do back in my day. This is one of those weeks. NVIDIA has announced that they are releasing the GPU kernel modules as open source. It's starting with compute, with some focus on CUDA and that kind of stuff, with an eventual solution for desktop graphics as well. And I think there's already some example code on their GitHub. This is going to change the game for distributing the NVIDIA driver on Linux. And we have a couple of really great links in the show notes for you to read up if you're curious about the details. And then also, I want to mention that Christian Schaller joined us for Linux Action News 240. And we dedicated the entire episode to our conversation with him to get some of the details.

And Christian is here now as well, live in our Mumble room. Christian, welcome back to the show. Thank you. Thanks a lot for having me. This must have been the hardest secret to keep there must have been a point where you realized this was really going to happen and that must have been thrilling yeah no for sure it's uh it's one of those things where i you know even it was even top secret inside red hat so i mean i had even conversations with other red hatters where like you're saying oh we realize you can do something here like yeah that would be nice

and because i couldn't confirm or this, what do you call it, not affirm it until it was public. So yeah, definitely a hard secret to keep for a long time. I bet. I mean, it wasn't too long ago that Matthew Miller was on Twitter saying, you know, this is something we've really got to address as a community. Getting the sense that a lot of people didn't know it was coming. And in part, I have to imagine this is something that they have to manage carefully because it is, as a company, it's kind of an indefinite commitment to now supporting something on an engineering level, isn't it?

Yeah. Now I mentioned Matthew's Twitter thing. Yeah, we ended up reading into the news after that. Think like, yeah, maybe we need to let Matthew know so he knows that maybe this isn't exactly the best time to stir the boat. That's fair enough. Probably a good idea, I suppose. So I'm curious. I know I got a couple of questions for you since we chatted. I suppose, as you'd expect, there has been some skepticism around the announcement. But I think probably the most prominent skepticism has been around this sort of line of thinking. Well, sure, they released the stuff that's going to touch the kernel as open source, but they just moved all this stuff to the firmware. And now there's a 34 megabyte firmware blob required.

And I've heard this from a couple of different people. I'm curious to know what your take is on that. Not like, you know, what the official position is but just your opinion on on this method of driver development okay so from a purely sort of ideal standpoint i would love for everything to be open including firmware that that said i mean i i guess the one reason i felt it was a little unfair when people started going after that was that everyone is doing this model of having a binary firmware and then you know the the driver being mostly sort of connecting that up. So, I mean, it's, yes, it's a little unfortunate, but it's

what everyone does. So this is not, I don't think it's right to single out NVIDIA. Then you have to basically go after more or less every vendor out there. I mean, the Linux firmware file is, you know, growing for every release due to more and more people putting more and more functionality into the firmware. So that's just the way it is. And it's sort of something we as a community have accepted, I guess, for a long time now.

Yeah, a lot of vendors are doing it this way. And I think the only thing that's a little unfair about the criticism is the file size is being used as like a negative, you know, 34 megabytes of firmware. But if you actually look at the firmware, there are just a ton of architectures that that 34 megabyte supports. In fact, Longhorn on Twitter did just that. He says the 30 megabytes supports multiple GPU generations, as well as a bunch of other ELF sections and things specifically for Ampere and Turing and all kinds of little bits in there that aren't necessarily like control mechanisms and stuff like that. So you can't really derive anything from that 34 megabyte file size. That's a bit

of a red herring. Additionally, it seems to me that that means the problem is solved on the Linux side. So now as a consumer, this is just a different set of pros and cons that you have to weigh. But it's clearly a benefit because it's going to make it easier for manufacturers like System76 that are shipping NVIDIA hardware. It's going to make it easier for folks that have existing NVIDIA hardware, but it's also going to be a benefit

for the Nuvoo driver too, which I think is sort of been missed in all of this. And I don't know, can you share any details there, Christian, about what's kind of maybe the next steps for the Nuvoo driver? Yeah, I mean, it opens up opportunities for us, right? So as I mentioned in my blog,

and it's that so historically what happened was that when NVIDIA switched to signed firmware, it basically meant that Nivo couldn't use that firmware. So we reached out and NVIDIA was kind enough to work with us to say, okay, let's make a firmware that lets Nivo keep working. working that of course ended up being a big effort for for nvidia because they had to maintain an extra firmware and it took time to get features in i mean one big feature people kept of course missing was the ability to reclock the gpu because nvidia gpus and i assume others by default default to sort of like almost like a power saving level of gpu so

so if you are stuck in that you know don't get a lot of performance out of your GPU. So we were in talks with them about like how do we fix this? And then, you know, when this thing became a thing, we said like, okay, maybe the solution is that we can then, you know, tie things together with your new release. And then we have a shared firmware between, you know, the NVIDIA stack and the MSS stack. Right. That is exciting.

That seems like massively simplification of engineering efforts and just way easier to maintain long term. Yeah, exactly. Casey points out in the IRC, too, that for a 34 megabyte firmware, you get a massively parallel computer with 12 gigs of RAM. Anyways, continue, Christian. I know you had another point there. Yeah, and I was just going to say with firmware, I think also i sort of came to terms with firmer in general when we launched the lvfs i felt that after that i couldn't really go out there and you know scream murder every time i saw former firmware out there being available close source i bet being involved with lvfs does give you a pretty good perspective on that i feel like this is one of those stories that after we

were done recording land i just wanted to take the rest of the week off. Like, I was so excited. It's like, what else could happen in Linux that would even matter this week? It's such a big deal. But now, really, in a weird way, now a lot of the hard work starts, right? Like, I look at the GitHub and I see all these issues that are being opened up by the community and people that are jumping in. And right now, we're really just looking at a solution for compute as well. And so we should probably set our expectations.

I mean, maybe it's not even unreasonable to expect maybe summer of next year before there's something people are going to want to install on their machine. What do you think? You think that's reasonable? Well, I mean, I think if you really want something to just run compute, like you want a driver and you want to sign it uh yourself for the time being to to run on let's say your blade machines to use it for compute then you're definitely ready to use it but you yeah you need to wait until the graphics comes in for one before you want to run it on your desktop and of course at the same time we in the distro world are of course making figuring out how we

want to deal with signing and packaging this thing and you know it's not an upstream drive yet that's another thing people need to remember it's still an auto tree driver so that has you know like any auto tree driver it's only a set of challenges for us in the distro world to deal with but you know i think the pieces will fall in place and it will make make things easier as we head going forward well this was you know really the result of years of investment in a relationship. And those details and more are in Linux Action News 240,

where Christian covers all of that. But I think this is a obvious win. I know some people are very skeptical, but to me, this means that people can now buy laptops or desktops within video hardware. And, you know know within a year's time it's going to work out of out of the box like your amd or intel graphics does that's just

indisputably a win if that means some some of the functions are in a blob yeah maybe it's not ideal but it seems to be the compromise we have decided to live with and it does mean we deliver on the best end user experience that we can right now. And it is such a tectonic shift from where we were a week ago that I think people just haven't really even processed what this means. So extremely excited. And Christian, thank you for giving us the scoop on all of that. And so we can help people understand what's going on and set their expectations accordingly. And again, I will link to your blog post in the show notes, which I think everyone should give a read over as well,

because you put a lot of great info in there. Thanks for joining us again. Yeah, my pleasure. And just maybe one last item. I would ask everyone to try to be patient with NVIDIA here too, because this is a big change for them too, right? And they have to come up with how they deal with the community around their driver. And I mean, NVIDIA have done open source before in various things. It's not

like they're completely new to it, but this is like the heart of their hardware, which is the heart of their business. So if things don't all go super smooth from day one, give them, have some patience and try to work with them as I iron out their own interaction with the community that will spring up

around the driver. Linode.com slash unplug. Go there to get $100 in 60-day credit on a new account, and it's a great way to support the show. Linode's been doing this for nearly 19 years, and their mission has remained unchanged since that began. Make cloud computing simple, affordable, and accessible to everyone. It's led to the best in class

experience. And that's why we host everything on Linode. We look at the options. I hear from some of you sometimes tell me about some great bargain basement deal you got. And then usually about six months to a year later, I hear about why you switched to Linode. So take advantage of the $100. It's just a great way to go try something. And Linode has a lot of one-click deployment applications as well. Some of the apps that we've talked about here on the show, one-click to get going. Zoom replacements, yeah, they got Jitsi up there. You want to try GitLab? Yeah, they got that. In fact, some of the services I'm going to deploy soon for some back-end stuff that I'll probably talk about in the future. I'm just going to use one of their one-click

deployments. They've also got a great one for WordPress, and they're always adding new ones just constantly. So I check in there from time to time, and I'll just deploy a project and try it out. And it's a great way to try out that $100 too. So from development tools to game servers to the backend infrastructure for your business. So I'm talking like scaling from a personal project to something for millions of users. Linode has you covered. And then they back it by the best support in the business. And they've got a bunch of nice features like S3 compatible object storage, cloud firewalls, DDoS protection, a powerful DNS manager. And of course, your info management too.

Kubernetes, Terraform, Ansible. They've got support for all of that. With pricing 30 to 50% cheaper than those duopoly hyperscalers that want to lock you into their crazy esoteric platform. Go use the real internet and go build a system either from the ground up or with one-click deployment. Linode makes it easy, and their community support runs deep. Go get $100, support the show, and try something out.

Linode.com slash unplugged. Support the show and try something out. Linode.com slash unplugged. Tailscale is a mesh VPN powered by WireGuard. We've talked about it on the show before, and I should disclose they are a sponsor on Coder Radio and Self Hosted. They don't sponsor this show, but you can go to Tailscale.com slash Coder or Tailscale.com slash Self Hosted if you want to sign up. And it's a tool that we all use as a team. And there's a lot of great tools out there. We've

mentioned Nebula in the past. We've also talked about S Shuttle or Shuttle in the past. We've covered all kinds of mesh VPN solutions like Tink and others. It's always a problem we've been trying to solve, but Tailscale really has seemed to solve it the best for us. And I had an opportunity to sit down with Avery Penneron, who is one of the co-founders of Tailscale. He spent the previous eight years at Google, where he launched some of their first P2P payment systems for Google Wallet. He was involved with Google Fiber and working out the Google Fiber Mesh Wi-Fi strategy. He's been involved in other high-profile alphabet projects. So it was just a great chance

doing these podcasts sometimes just gives me a chance to talk to people I normally would not get to talk to. And I wanted to get kind of an insight into tailscale because it just almost seemed too good to be true. And I was hoping maybe to get some insights. And I walked away from this conversation with some of those, I think. So I wanted to share it with you. Avery sat down with me just recently, and we were chatting just before we got started about some of his previous projects that I was familiar with. I have been looking to solve my disparate network problems for years, and I've been rolling my own crazy solutions. I had tried a few different things with WireGuard, but I actually started with Shuttle.

And we talked about it a long time ago on the show. And I believe you were the original creator of Shuttle VPN, correct? That I was. 2009, I think. So you've been thinking about VPN stuff for a while, I take it. Was WireGuard like the missing piece for you? WireGuard was definitely a missing piece. I've actually been thinking about VPNs all the way back to 1997 when I wrote my first one. WireGuard like the missing piece for you? WireGuard was definitely a missing piece.

I've actually been thinking about VPNs all the way back to 1997 when I wrote my first one for the first startup that me and my roommate started in our dorm room in university. And that particular one made some major structural mistakes that you should not make when designing a VPN. We didn't exactly roll our own crypto, but we took the crypto primitives and we put them together in ways you probably shouldn't do. I learned a few lessons from that. We did TCP over TCP, which you should not do in a VPN. But it was educational. And in fact, this carried our product forward. We had a company, it was the thing that actually launched us, this VPN, because it was on a network appliance server you could install in small businesses. And it was quite easy to configure compared to other VPNs at the time.

But it had these other flaws. Fast forward to the shuttle days. I'm like, OK, I learned a lot of lessons. I know a few things not to do. And I'm a little frustrated that our product got acquired by IBM. IBM sort of dismantled it and killed it. No more VPN that I liked. Alternatives are like PBDP and IPSec, which are pretty terrible, or OpenVPN.

So Shuttle, the motivation was like, I don't want to do the encryption myself, and I don't want to do TCP over TCP. So SSH has good encryption. Let me just like piggyback on top of SSH. You know, rather than sending the packets over SSH, I'm going to disassemble the TCP sessions and then reassemble them on the other side,

which bypasses this TCP over TCP problem. That actually worked quite well. And it was an open source project. I didn't try to commercialize it, but I used it for all kinds of things. It helped, you know, any server that supports SSH can be your VPN server, which is quite handy. The performance wasn't the greatest, but it was perfectly good for like transferring some files around and doing some remote terminals. With Tailscale, when WireGuard showed up, I was really excited because it is like, look, I don't have to do my own encryption, but I can understand the encryption that these people are doing, which means I can take this simple primitive for encryption

and plug it into a bigger picture of the way I want the world to work. A really simple VPN that I control everything about the way the experience happens, but I'm not sacrificing security. It is deceptively simple on the front end. As a customer experience, it's a very simple application to install. It's very simple to authenticate. It uses my existing single sign-on provider. It really couldn't be smoother. And so I'm not surprised to hear that you've been thinking about this since 97, because when I first tried Tailscale, I thought, they must be pulling some real magic off on the back end to make this feel so simple.

So what are some of the more complicated things that are being solved on the back end to make it just such a seamless thing for me to have a flat network of all my devices? There's quite a bit of magic in there. The first step is just this key management. So in Shuttle, we got out of key management by just saying like, look, it's your job to do SSH key management. You already know how to do that. We'll just piggyback on top of that, which is great,

except that in fact, many people do not know how to do SSH key management. And so there's a bit of a limited market. It's only people who can figure out how to manage their SSH keys effectively, which at least is most developers. If you wanted to deploy that company-wide, for example, most of the people in your company are not developers. And so they don't know how to set up SSH keys. And so Shuttle was never a useful product for the general community, even if pushed by an IT team. So Tailscale's key management, instead of using SSH keys, we used what is now, you know, 12 years later, 13 years

later, keep losing count, 13 years later, really common, which is the single sign on system that people are using on web apps, right? Click, click here to log in with Google, click here to sign in with GitHub. This stuff in the intervening years has become like the standard thing to do. Everything is turned into a web app. All the web apps have a button for like login with Google. So we make Tailscale look like a web app for the purposes of authentication, right? You load up the Tailscale app, which is a native app on your phone or your PC or your Linux machine. And it gives you a URL, which sends you to a web browser, which then acts like a SaaS app, which can use your single sign-on. So that's the first trick. We piggyback again on somebody else's authentication, just like Shuttle did.

But this time we're piggybacking on authentication that everybody knows how to use. The second thing we did is what we call NAT traversal. So this is especially magical. Other VPNs I've built myself and have used require one end of the VPN connection to be on a public-facing open firewall port, which is not that shocking. The whole internet kind of runs on this principle. The client can be behind a NAT or a firewall, but the server has to be outside waiting for your connection. Otherwise, how is it going to work? The problem is people setting up a VPN

or trying to make a connection between two devices, this is really limiting to what kind of connections you can make. It essentially turns the whole internet into this centralized system, where nowadays the only people that get a public IP address are cloud providers, and then you have to rent one from a cloud provider, and now everything is like the hub is the cloud provider, and the spokes are everybody else, right? And you're paying rent to these cloud providers

to do anything you want to do on the internet. So Tailscale, the magic involves like you could literally have you in a cafe with your laptop and your coworker in a cafe with their laptop. They log into Tailscale on the same domain. Those two devices will find each other. They will not go through an intermediate server. They'll actually make a direct peer-to-peer connection between those two devices. and now they're connected. You can go to a service that your co-worker is running on their laptop and try out that web server, so you can SSH into their device or do file sharing or whatever. Well, this is sort of what I've discovered using Tailscale, is that it is essentially allowing me to create my own decentralized internet, and it's kind of a sneaky way to do it.

It's not really like Tailscale selling itself as build your own decentralized internet. But that's kind of the end result, isn't it? Well, I mean, the word decentralized has become a bit dirty now, because every time someone says decentralized, they assume you're talking about blockchains and bitcoins and stuff. Sure. Yeah. Yeah. Or any kind of weird crypto. Yeah. And that's not what Tailscale does. The way I like to think of it is like, look, we actually did decentralization in a way that's actually decentralized. Therefore, it achieves engineering goals, right?

We're not doing decentralization for religious reasons. We're doing decentralization because it's the most efficient way to solve your problem. Well, and it really is about connecting my devices, isn't it? That's the issue is I don't have a great way to connect all of these disparate devices, double NAT, different cloud providers, VPSs, some of them are locked behind VMs or containers. It flattens all of that. So the neat thing about Tailscale is that it's sort of what I call a hybrid centralized decentralized system. The control plane, the part that decides which devices can talk to which other devices, the part that sends your public keys around in the same way that you would send a public key to an SSH server and put it in your authorized keys file. That part is centralized. You go to the tailscale SaaS product.

We run a little tiny thing that we call the control server that is basically just a collector for public keys, but also connects to your identity system, right? That is where everybody else goes to rendezvous. So me with my laptop in this cafe, you with your laptop in the other cafe, both of our laptops contact the control server and say like, hey, where are my friends, right? And you find out the IP address and the public key, go back to that. And then those two laptops do the data plane, actually sending packets back and forth, initiating connections with each other in a purely decentralized way. This ends up being a really elegant algorithm because it's extremely inexpensive to run a centralized control plane where they only use it occasionally

and there's no latency sensitivity. But it's extremely inexpensive to run a data plane where we don't have to transfer any of the bytes for you. Right. Keeps the infrastructure cost down. In fact, you had a blog post that I really liked. It's titled How Our Free Plans Stay Free. And you write in there, All this is to say our costs are carefully managed. Like other SaaS companies, we don't build physical infrastructure. We avoid touching your packets for privacy, but also to reduce our costs. We fix bugs and docks instead of answering the same questions over and over again.

Our control plane is lightweight, and our DEP network is cost-controlled. This allows us to maintain a healthy operating margin so that a free tier isn't competing for resources with our paying customers. And I think there's a lot of insights just in that paragraph right there that I felt like I got about the company and about you. I mean, first of all, it's very clearly, plainly written. It's very easy to understand. You don't really mince any words there. And it makes a lot of sense. As a customer, my first concern was,

well, is this thing going to stick around? And is that a problem that you have discovered? Are customers worried that this thing is too good to actually stick around? That happens pretty often. One of the most common bits of feedback we get is like, please, please don't cancel the free plan. Please, please don't get acquired by a list of competitors.

Please don't add any more features to the product because I like it just the way it is. And I know what happens when companies add features to their products. I feel like the whole industry at this point has kind of been scarred by a bunch of software companies that have been, I guess, kind of selfish. They care more about their business model than they care about their users. It feels like the VC model in particular is tricky. And so I wonder if we could speak to that for a moment, because also a congratulations. I know you recently raised a Series B round of funding, and it looks like you have a blog post that addresses that as well.

And in here, you know, you kind of address that issue is like some people are concerned after they hear about some series funding that there's going to be a whole bunch of feature creep that all of a sudden it becomes this giant platform that like a chat system has to be built on top of. What are your thoughts around that? The biggest concern that people and even me when I see announcements like this have is if you had to raise money, does that mean you were going to go to business if you hadn't raised money? Because if that's the case, then you're on what we call the VC treadmill, where you have to keep running faster and faster just to not fly off the treadmill and smack into the

back wall. Most startups find themselves in that situation. And when that happens, the desperation level rises and rises. And so you'll see a product that starts off really good, but they didn't quite find their footing, couldn't figure out quite the right business model, didn't make quite enough money to pay their expenses. And they had this tough decision, am I going to lay people off? Or am I going to raise more money and keep trying? And most of the time, the companies go and try to raise more money and keep trying. But that, you know, they keep trying and it may not ever find

their way, right? And that's where you get these more and more desperate, weird things where like, I'm going to add a chat system, or a chat bot to my unrelated product. Tailskill is a little different. And I tried to get that across in this blog post. So we announced a series B funding of $100 million, right, which is pretty big for a series B. We're not raising that money just to stay in business. We actually had a bunch of money still left over from our series A. We were on track to be profitable by the end of 2022, kind of inefficiently, because in fact, it's all, you know, if you're going to raise this much money, you should actually spend it. Otherwise, you've kind of wasted your equity selling it to investors. So we intend to spend $100 million. But we want to do this in a way that we never end up at that desperation level. Like we want to be the ones in control. We're

tracking our spending rate related to our revenues. And because we've got extra money, we're trying to sort of clear out this buffer of extra money by spending a little faster than our revenues. But the revenues, we look carefully at how fast we're growing. So we could become profitable at any time we want. And that's a really important attribute to maintain in a company where you don't want to end up at this level of desperation. So we believe we can do it. I mean, you know, there's famous last words, we can always make mistakes. But we really, really put a high premium on not screwing this stuff up. Because all three co founders, we're getting a little older now we're into our 40s.

We've seen this before, we don't need to make a million dollars tomorrow, but we are really tired of spending five or seven years building something and then it fails and gets canceled out from underneath us or acquired and then somebody, you know, tears it apart, right? I don't want that to be my legacy as a person. When I retire, every project I've ever done has been ripped out from under me and torn apart. I've noticed what seems to be a consistent long-term thinking in a lot of this. In a lot of what you communicate, you're currently thinking about what we're doing today, but you're also thinking about where things are going long-term. And I wonder, is there something particular in your past experience that has really drilled this into, we have to be careful when it

comes to VC funding and we have to think in a certain way? Is there something that you went through, an experience you went through that really codified that? I mean, my first startup, of course, was in this category. We absolutely got ourselves into this trouble where we built some really cool technology, including that VPN I mentioned, the pre-shuttle one. It also had a bunch of other stuff. We found the traction. Users were getting excited. We never quite figured out our go-to-market model. So we could sell the product, but the expense of having salespeople calling people to try to find the ones who were willing to buy it was always a little bit higher than the actual revenue we got from selling the product, right? And this is in the pre-software

as a service days. So it was kind of weird back then you would sell something once and not a monthly subscription. And that, you know, the whole mathematical model of that is very different from the monthly sales, right? So if you have a sales team today, each salesperson, say, can add 50k of ARR per month, right? Then that's great. You just have this one salesperson and they keep going and the ARR keeps going up. And of course, you're going to have churn and so on. So sooner or later, it maxes out. But you have this really nice model. When you have people selling one shot each time, then your revenue doesn't keep going up over time, right? The salesperson that you have selling a certain amount per month

means you get that much revenue per month, not that much additional revenue per month. So the models are very different. But back then, we were selling appliances. You bought appliances. The device was, there was no subscription element to it. And so we got ourselves into this treadmill of the long term value of a customer was slightly less than the cost of acquisition of a customer. And eventually, we ran out of money related to the.com crash and everything else. Or I shouldn't say we ran out, we never actually ran out of

money, we ran low on money. And eventually, we had to look for an exit strategy. The exit strategy we find was through acquisition, right? So our investors made their money back, but none of us got super rich from it, right? And then I worked at Google for a while. I don't want to say too much about Google because everybody talks about Google all the time, but I saw many, many projects inside Google succeeding and failing for many, many different kinds of reasons. And I'm extra curious about that stuff. So I talked to a lot of people while I was there to understand what was working in their

project and not working in their project. Like there's definitely patterns to the things that work and don't work. And a lot of it involves just sort of like, I think I'm not, I'm not a real sports person, but some people call it like getting ahead of your skis, right? When you overinvest money and start spending faster than you understand the business model, that's when you get yourself on this treadmill. And it happens just as much in startups as it does in teams and large companies. Do you feel like larger teams are more susceptible to that problem than smaller teams and smaller companies? It's not so much about team size, it's about money, right? So the curse is having too much money and not having the self-control to know that

it's too much money. Back in the old dot-com days, people used to say this about Amazon, losing money on every sale, but making up in volume, right? A lot of startups to this day still try to do that, and it still doesn't work. Amazon was not actually doing that. Amazon was doing something else, which is why Amazon is so successful. But it kind of looked like that. And there were a lot of startups that were doing that at the time, but they still do it now.

If you have a business model that you can't make money per unit at a small scale, you're definitely not going to make money per unit at 10 or 100 times that scale. I want to shift gears, but kind of in the whole business model arena, Tailscale is open source. And most of the times when I talk to CEOs, they tell me that they can't open source their software because somebody would just copy it, stand up a clone and take all of their profits. And that doesn't seem to be a concern of yours. I mean, it's a concern. It is. It's also a backup plan. So like I said, one of my number one concerns is like when I retire, I want my stuff to still exist. And one of my favorite things about Shuttle, which is actually one of our competitors, is that no matter how much I lose attention in it, no matter how much money it doesn't make, Shuttle is still around, right? There's no company that can just disappear and vaporize the software that was attached to it. And I really like that about open source. I think there's a lot of people out there who open source things, and that's like

at least half the reason they do it is so that when the company they're working for goes away, or when they leave the company they're working for, they can still have access to all that source code that they wrote is not destroyed, right? So I have to admit, I have an element of that. Like, I personally want to make sure this thing can never die. And this is one way to do it. That said, it's also beneficial to our customers. First of all, you can get bug fixes and contributions. Secondly, people don't like installing proprietary VPN clients in their Linux machines. Just generally speaking, you've got almost the whole stack you're using in your cloud servers is open source. If you have this one proprietary VPN server, it very much slows down the adoption.

And thirdly, there is a component of Tailscale called the control server that we operate that is not open source. Now, there is an open source control server called Headscale. It's not maintained by us, but we contribute to the Headscale development. We work with the developer and so on. It doesn't have as many features as Tailscale. It doesn't integrate in exactly the same way, but it works with the same Tailscale clients. So people who are really concerned, they want to have 100% open source system can do that. Yeah, I was going to ask you about head scale. So I'm glad you touched on that. Thank you.

Looking at tail scale, I think a lot of people identify it as one of those quintessential, this is how I get my work done now tools. And there's others out there, Slack and Dropbox come to mind. Do you watch some of those companies or are there companies that you watch and go, okay, we're not going to do that? What are you looking for not doing? Like, I'm curious as a customer, like, what are you not going to do? That's an interesting question. In the case of Slack, I really like the way they've done their business model. The way they did their authentication was a series of missteps that I think they're

finally getting to correcting many years into their existence. So we're probably not going to copy the way Slack did authentication. But I really love their business model, right? The way that individuals in a company can adopt Slack. They don't even need to necessarily know that other individuals in the same company are adopting Slack in different Slack instances. And then at some point, you can group all those instances together and say, okay, we're all using Slack here at this company. Why don't we get by a company subscription and get a discount, right? That is a really nice business model. That's something that in the security world and

infrastructure, which is what where a tail scale finds itself, you almost never see that like there's almost no such thing as a security product that is adopted in this bottom up way. And the reason for that is pretty simple. It's that security products almost never solve the problem you have right now, right? If you're a software developer, you're trying to get something done. The only thing a security product tends to do is stop you from getting that thing done. That's the job of a security product. So it tends to be enforced top down by the CISO saying like, look, I know you guys have a lot of work to do, but could you please not do these five things that are horrific

security holes that are going to get us in trouble? And then they like slow you down. Tailscale has this nice model where it comes in as an infrastructure tool, a problem solving tool. And then coincidentally on the side, it's also really secure. So the default thing you do with Tailscale is way better than the default thing you would have done otherwise. So the security people who are in charge find out you're using Tailscale. And instead of saying, oh, no, no, please stop doing that. They're like, oh, thank goodness you stopped doing that thing you were doing before. Let's maybe formalize this Tailscale thing a bit. It makes sense. It does definitely feel like a

bottom-up tool that people are discovering and just falling in love with. I'd like to kind of wrap it up by just kind of getting your thoughts on sort of the state of the tech ecosystem. You know, you look out there, we have some major, major tech companies, and it feels like eventually, if you're playing on their platforms, you get their attention. And I'm wondering if that's on your radar at all, as far as maybe Apple goes or Microsoft or Google goes. Have you shown up on their radar yet? Have you had conversations with them? Are you concerned at all about them influencing or trying to influence the direction of features on the mobile app or something in that regard? I mean, we've had interactions with pretty much all of them. I would say that in general, they're all pretty happy that we exist. We're not hurting anybody's platform by creating this

functionality. There's no like embrace and extend model for them. Because I mean, if you imagine Apple, for example, there actually was a tailscale like product that Apple launched quite a few years ago now called back to my Mac, right? It did almost that exact same thing, except it only worked Imagine Apple, for example. There actually was a Tailscale-like product that Apple launched quite a few years ago now called Back to My Mac, right? It did almost that exact same thing, except it only worked on Apple products, right? Which sounds neat, but almost never solves your problem, right? And similarly, Microsoft, I just found out yesterday, there was a product called Windows Mesh, I think, came out apparently with Windows 7 or near the time of Windows 7 that did similar things to Tailscale, but it only worked on Windows.

Tailscale has a feature called TailDrop, which is a lot like AirDrop that Apple has that lets you move files between your devices, except Tailscale's one works on all your kinds of devices and it doesn't care about physical proximity. Apple's AirDrop only works on Apple devices, and even then only sometimes, and nobody's quite sure why. It doesn't always work, right? And AWS, for example, they've got an AWS VPN, they've got VPCs, they've got IPSec connections, but it only works inside AWS. So if AWS was going to try to make a tailscale, it might not make it easy to connect to other

kinds of clouds, because they really have no incentive to make it easy to connect to other kinds of clouds, right? But if tailscale is sitting out here independently making all these things work, then it actually makes it easier for you to use AWS, easier for you to use Windows, easier for you to use Linux, and easier for you to use Apple. It connects all these things together. It makes all these platforms better. So we don't see a lot of interference from the platform owners because we're not really hurting their business at all. We're just making it so that their thing

works better with everybody else's things. Yeah, really. It makes it so I can get more work done on their devices. It's what it does. Exactly. Well, Avery, I'm just a huge fan. So thank you very much for all of the hard work and of course for everyone there at the team.

And thanks so much for coming on and chatting with us. Yeah, thank you. I really enjoyed that chat with Avery. it's not many founders i can sit down and talk everything from tcpip to uh big tech economics so that was really great and again if you want to sign up that wasn't sponsored i wasn't paid for but if you want to support one of our shows tailscale.com slash self-hosted or tailscale.com slash coder. The whole idea of mesh VPNs that can defeat double net carrier grade networking has been a topic on this show for a long time. And when we were in Raleigh, I was kind of leaning over to Brent and I was saying, you know, you got to try this tailscale thing out. I think you would really like it specifically

because you do kind of travel around and you've kind of got places you visit frequently now. Tailscale feels like it could be a good solution for you. And I know you didn't get a chance to try it out while I was there, but I think you have sense then. So what is like the, I don't know, let's say the Brent review of Tailscale. What are your thoughts? Well, I got to say I'm relatively new to it. It's been maybe a week that I've been into it and maybe a day since I've really dove into it on my own systems here. But whenever you nudge me towards something, I know I need to be listening. And so thank you for that. I have in the past, you know, with travel in mind, set up a wire guard exit node. So I have a little VPS

somewhere that just purely accepts my connections wherever I happen to be. That makes me feel a little bit better because there are certain places that I go with my laptop that I don't feel like open traffic is necessarily a good thing. But that being said, now I have access to a VM that Alex set up for me where I've been doing some Ansible playing with. And I guess I also have a Bitcoin note sitting over there that I can access from here now. And so the more I dive into it, the more I realize it's solving a lot of my various issues that I have had in the past and have found solutions that were a bit annoying, but also solutions for problems that I've had that I haven't had a solution for yet. So I know, Chris, you use it a whole bunch, but I'm like tingling with excitement now

of diving into Tailscale and really getting a sense of how it can just make things easy and seamless for me. For me, the progression was at first for like a month or two, I set it up and I was like, oh yeah, finally I can connect my RV server systems and the studio systems all on one flat network. That was a huge thing for me because until Tailscale, I never opened up the RV's network to the internet. And there's a ton of like dashboards and control panels and statistics and data that the system in the RV is collecting all the time.

And sometimes I'm at the studio and I want to check in on that stuff. So that was the first problem I solved, right? Because it takes 30 seconds to get installed on a Linux box, if that. And they have Raspberry Pi packages and stuff. So I was like, all right, this is great. Yeah, okay, sweet. I got a VPN. All right. I didn't really get hyped about it. I don't know why it took me like three months. So you're way ahead of me, dude. Until I realized I could install it in a VM and I had several VMs and projects, like a lot of the

Knicks stuff, the Knicks challenge stuff I was doing in VMs. And it dawned on me that, wait a minute, I could set up a Nix VM, install Tailscale, which is awesome on Nix. And then I could access this VM from literally all my machines, like it's right here. And I could continue the next challenge from anywhere, which was great because we were traveling to North Carolina and all that. And that's when I realized, oh, wait a minute. I actually could run it in this container. Oh, and look at this. Umbral has a tail scale app that runs in Docker.

And now I can get access to my Bitcoin. And I just. That is cool. Yeah. That's when it clicked when I was like, oh, I could do it in VMs. I can do it in containers and I can do it on my physical devices, including my mobile devices. And so I just sort of built out a bigger and bigger network. And then I kind of clicked, oh, I can also, you know, support my kids computers this way. So I really like it. I've probably crossed over where I have more systems using tailscale than Nebula. Nebula is great,

because there's like, no centralized service, there's nobody getting involved, but it's a lot more of a bring up. So it just didn't quite hit the sweet spot for my like my laptop and my phone and my Raspberry Pi. But it was it seemed like a really sweet spot for the VPS and the servers on the studio land. But now Tailscale, I think, has kind of eclipsed some of that, too. And it has some really nice DNS stuff. You can actually, if you wanted to do an easy mode, you can put a Pi hole in your Tailscale, I think, has kind of eclipsed some of that, too. And it has some really nice DNS stuff. You can actually, if you wanted to, do an easy mode. You can put a pie hole in your Tailscale network, and it will start resolving DNS for all of your machines on Tailscale. So you can start referring to all your machines by machine name.

That is very cool. Yeah, I like it. It's also been really handy while I'm on the road again. Every freaking time I'm on the road, that's how I, a lot of this, a lot of like the, the boosts and feedback. Uh, I do all of that through a VPN. It's really nice to be able to just have that right there. So it's a, it's a hell of a thing. I'm really impressed with the service and that's why I was hoping to get some insights into how the heck they're going to keep it running. And the, the thing that he touched on in there that I liked is they've set it up in a way where the free accounts don't compete with the paid pro users, which is always like this collision that it seems like all of these services that offer a free version are always heading towards this iceberg where inevitably the sales team feels like the free accounts are competing with their benchmarks.

And, you know, we end up you always see like the free stuff getting cut back. And the idea that they've kind of designed it in a way where that doesn't happen. I feel like that shows some really good documentation just to get a sense of how it all works and what I could possibly do with it. And I found their documentation to actually be really, really great and also fascinating, which I don't say very often about documentation. It's slightly, I'll say it's ever so slightly confusing. They have a section for documentation, which I think is a little bit more on the technical. It's a simple explainer of the technical side of things, which I found very, very helpful. So getting DNS set up for your whole network and those kind of useful tips. But they also have sections that I would encourage people to just go browse. One of them is called the Tailscale Guides, and the other is Tailscale Solutions. I don't quite know the difference between them. They seem very similar, but there's some amazingly fascinating stuff in there. So they have a guide on like accessing a pie hole or raspberry pie from anywhere, which I think is a really nice place to start. Maybe if you are trying to play with this but they also have you know a guide on tailscale on xos a new minecraft server in 10 minutes which sounds really fun tailscale in

lxc containers i know we talked about that a little bit last week so that might be a place to go uh if people are interested and the last one that really caught my eye was um setting up a dog cam with Tailscale Raspberry Pi in motion. Bitwarden.com slash Linux. Get started with a free trial of Teams or Enterprise or for an individual when you go to Bitwarden.com slash Linux. Simply put, it's the easiest way for individuals or businesses to store, share, and sync sensitive data. Bitwarden is fully customizable. You can turn certain features off that you don't want.

And of course, it's what I use for my password and secret information storing. And I'll also add my two-factor authentication token. I put it all in one package. And man, is that handy when I'm traveling. Oh my gosh, you guys. I won't name names, but some of my friends out there, they don't use Bitwarden.

And I have been on a campaign to convert them because it is painful when I sit down and watch them try to log in on some service when they don't have a password manager. Like maybe they set up a new box or maybe they're using a system in the studio temporarily. Oh, my God, it's painful. Bitwarden makes it so much simpler. studio temporarily. Oh my god, it's painful. Bitwarden makes it so much simpler. They make it easy and safe to store sensitive data for yourself or when you're working with the team. Wes and I use it to manage our passwords and two-factor codes and other sensitive data all the time. I use it to work with team members, and I really think it'd be great for open source projects too. And one of the things they've just recently added, which is brilliant, is a username generator. Stop using the same username at every site you go to. I never really thought about it, but that's crazy. Anything you can do to improve your security online is going to be good, right? But like, I'd say number one is passwords. And number two, I never really thought about it, but of course, it's usernames.

I never really thought about it, but of course it's usernames. Because if a service is compromised and their database leaks, then that username is going to be the same username across multiple sites. At least it is for me. I'm embarrassed to say it's true. So having a different username for each website you use, that's going to make it difficult, I'd say impossible, for a hacker and attacker to track you across sites. And of course, their generator supports your email service

if it has like the plus edition for an email. So like Gmail, for example, user plus your actual like fake thing at gmail.com or whatever. It's really nice. And it's so intuitive the way they do it because it all just stores in Bitwarden. Then when I try to log into a service on my phone, I have a nice secure password for that, too. And I'll admit before they got really good password management on the phones, there just wasn't a great option. And Bitwarden nails it. And then on the iPhone, it ties in with Face ID.

So it's scanning my face. It's importing the password. I mean, it just happens so smooth, so flawlessly. I love it. And like I say, when I'm traveling tools like this, I don't know. I don't know how I got work done without them. Something like Bitwarden, it's the only way I'm going to go.

And I love that it's open source. And I love that they have a self-hosted option. And I love that they got a giant community. Go try it out for yourself, or maybe recommend it to a friend and family member like I've been doing recently or maybe it's time for your workplace to kind of get their act together in this regard. No shame in that game. It's just, you know, the way things have been in the past but now Bitwarden's changing it and they've got great team plans too. Individual or team, go try it out right now. Bitwarden.com slash Linux. Go to Bitwarden.com slash Linux to support

the show and to get started with the best password management. And I'll say it my personal way to keep all my secrets. What's in there? Bitwarden.com slash Linux. We got some great feedback again this week. Thank you, everyone. Lyndon wrote in, I was just listening to episode 457. I'm a heavy user of Ansible and LXD and thought I'd share a couple of useful tips. One is the ability to use LXD to manage hosts with Ansible. No SSH required. If you configure a local LXD client to connect to a remote LXD server, you can specify the name of the host and containers in your inventory and

configure the Ansible host variable as follows. It gives an example we'll include. I think that's kind of neat. I know Alex got real excited about that. So that's sort of fun. They continue. I know you folks are also Bitwarden users. Rather than store secrets in multiple places and use Ansible vaults, I found a plugin that lets you use the Bitwarden command line client to retrieve Ansible secrets. So we'll leave a link to that. That is awesome. I didn't really even think of that. It's so great. Why not use Bitwarden for the secrets? You know, what he touches on here about being able to deploy a LexD container or a server with just like a couple of things that you specify and then ansible does the rest i do really appreciate that when we

did the nix os challenge we did a bit of a comparison between ansible and nix and that's sort of what's brought up all this ansible feedback and it wasn't really a fair comparison because you know the nix system lets you manage the nix package and the nix os possibly but you could run the nix package manager on just about any distro and it lets you manage the Nix package and the Nix OS possibly, but you could run the Nix package manager on just about any distro. And it lets you manage that world, right? Where Ansible is like multi-OS, multi-distro, multi-container platforms, right? And so that's the big difference. Now, I think there is a totally valid argument for only needing to learn a Nix system if that's the scope of your work, but I think it's also really valid to say, you know, there's a lot more I want to get done. And so Ansible is the way for me. And I totally appreciate that. And I feel like I've gotten a whole new set of insights on how people use these

tools accidentally. Like we just stumbled into this. And I'm learning a lot. So I guess that's a good thing, right? So yeah, and I think that's the really the beauty of some of the feedback we keep getting. We're in a position where people just keep giving us some great tips on things we didn't know we needed. And it turns out we do. And now it is time for the boost. As always, we get some great boosts as well. Sir Lurks-a-Lot is back. 1337-SAT seems oddly specific.

Feather for the tinfoil hat, if you will. Starlink knows where you are, even if you bug out in the middle of nowhere. Considering how Elon's car company collects an enormous amount of data about its customers, you have to at least wonder how that data is being used, or sold, or in what form, who actually has access, and how securely it really is in the hands of yet another tech company that dabbles in hardware. Are your internet habits and travels valuable to hacker loot? Hmm, maybe some bacon here, Chris. Oh, man. Yeah, I hadn't really thought about the fact that now there's one more company that's tracking my exact location and they do you know

they have to know where the dish is in order for it to connect to the satellites up in space and the dish is always aware of its longitude and latitude and um i must have gps in there too you know it's such a trade-off um i don't know why but i probably am more comfortable with starlink than i am verizon or at&T or T-Mobile. I feel like Verizon, AT&T, T-Mobile probably have like an automated system that law enforcement can just check a box and get your location info for like the last year. Maybe Starlink will do the same, but it's the trade-off of getting connected to the worldwide community, right? How else can you get connected to an internet like that

without some trace unless you use someone else's system and joops does have on the roof i have a seven and one antenna array and a couple of those antennas are dedicated to 2.4 and 5 gigahertz wi-fi i can pick up wi-fi from quite a range so So every now and then there is, I do have the capability of doing like some sneaky network joining. Careful what you admit to here. I know, I'm sort of realizing right now, but you know, when you're out in the middle of nowhere and you have to bring your own internet, yeah, that would be a problem. Marcel wrote in with a thousand sats. So when does the Star Trek podcast come out? I'm kidding, of course, but I would definitely listen to that. You seem very passionate about it. i just finished enterprise and i have mixed

feelings on the pre-show for the member feed i think we got into star trek again didn't we i think it was all coda radio's fault it happens from time to time we get up but with age does come some wisdom and i have realized that there are some hobbies and passions that I have that I shouldn't make content out of, and I should keep them for myself to enjoy. And I've decided that Star Trek is one of those things. And I have dabbled in the past with Star Trek related content. And I have found that Jupiter Broadcasting is its strongest when we really stay on mission and focused. And so that gives me an opportunity. That is an excuse to just keep Star Trek for something I personally enjoy.

However, I do love a good Star Trek conversation. And I will always jump in if something's going down on stream related to Star Trek. And I would love to, like, if one of my friends out there starts a Star Trek podcast, I'd love to join them for an episode sometime because, yeah, I'm a bit of a fiend when it comes to Star Trek. But I don't plan to launch a Star Trek show. I think I'll leave that out for all the other folks out there that want to make podcasts around it. We got several just like thanks for doing a show, Boos. These are great, too. We got a couple from Crashmaster18 and Wine bear in a total of about 6 000 sats they

just sent in some thank yous so i just want to say thank you guys for boosting we also got some folks that are doing the sat streaming so you can set an amount and while you listen it just it'll stream a sat every now and then back to us and the great news is is that with the price of bitcoin down sats are on sale they're the sameats. One sat still equals one sat. So get them cheap while you can right now. I'll probably be on sale for a little while as the market's going to just be in the tank. So go get some cheap sats and load up a podcasting 2.0 app. You can get them at newpodcastapps.com. Podcasting 2.0 has a lot going on, including they're working right now very diligently on incorporating live audio streaming into the app. So you just open up your podcast app and if Linux Unplugged

is live, it'll just show as in your feed list, it's live on air and you can tap it and you can listen to our live stream in your podcasting 2.0 compatible app. Not all apps support it yet, like I don't think fountain or castomatic do, but some of the web apps, there's some really good web podcasting apps, by the way, who knew podverse is great because it's a web app that syncs to a mobile client as well. All of these are listed at new podcast apps.com, but I can't wait for live streaming and transcription and a whole bunch of other stuff. That's all part of the podcasting 2.0 spec along with the value block. You can check out a lot,

all that info in office hours. We covered a bunch of it last week at office hours dot hair for that. And go try out a new podcast app, a new podcast apps.com. And if you don't want a new podcast app, because I know a lot of you love antenna pod, that's very popular in our community. So is overcast on the iOS side.

And actually, Apple Podcast app is always a strong contender. And Pocket Cast, of course. I know you guys love those apps, and you don't want to change. Go try Breeze. You can send in some boosts via the Breeze app, and you don't have to switch podcast apps. That's pretty, pretty nice. Guess what?

We got a pick. And it's on theme. Like I likes it. Brent, how do you think you pronounce this one? Oh, I think I got this one. You ready? You do? Oh, yeah.

Go for it. I think it's called We're On. Ah, yeah. Yeah, like we are on. Right? You see that? Yeah. We're on.

I think you nailed it. So this is a secure overlay network based on WebRTC. That is interesting. Yeah. This is weird. But it's on theme for this episode. Wes found this, actually. So it's a shame he can't be here.

So maybe if we ever give this a go with him, we'll bring it up again because I think he actually wanted to talk about it. But the idea is with Waron is that it gives you access to nodes behind a firewall, behind a NAT, using WebRTC to establish the connection between the nodes. And actually, this stuff has been built into WebRTC all the time, and nobody ever built the tooling around it. And it always drove me crazy because we basically use WebRTC now for video calls and audio calls. And that's really where we stopped. But you can actually build a freaking CDN on top of WebRTC. So they're using the underlying suite of tools that are available for this. And the project claims it's pretty low overhead and it's pretty surprisingly good performance. That's ridiculous.

If you use it between nodes on a land, they say you may not even notice a performance overhead, which is just wild. I don't know, man. I don't know, but it's actually pretty easy to get up and going on a Linux box. You basically can just pull down a binary and get it up and going. So we'll put a link in the show notes to where're on. And I could see someone who wants a temporary mesh network. You know, maybe it's up for the day, or maybe it's up for an event. I could totally see that. Oh, yeah, right. You're doing a corporate event or a Linux event, and you want to build a temporary mesh network. You don't necessarily want it assigned

to anyone's particular account or anything like that. You could use we're on to stand that thing up in five minutes. And now you got yourself an overlay network all running off of WebRTC, which everybody has support for on their box. Not really something we need because we're pretty much set. But gosh, you can see that. You never know. That's it. That's it.

We're wrapping it up. I just want to make a mention that we do this show live on Sunday, even if that means I got to set up some sort of hack together, set up run over cellular from Southern Oregon, we're doing it live. And that means that we get great folks like Christian from Red Hat that stopped by. And if you're in our if you're in our mumble room, and you got a question about this NVIDIA open source driver, you can ask one of the guys that has the most knowledge about the situation directly. That's a unique opportunity. And it happens right here on Linux Unplugged. But also hanging out live just lets you listen in real time, give us feedback in the chat room.

And if you get Mumble, you can actually just hang out in the quiet listening room. Then you get an Opus codec quality of the show with probably lower latency than Brent and I have between each other right now. It's great and it's a totally free software stack. And then you got it. You can join us and you can do things like office hours, which we also do on Mumble and other network shows that use Mumble from time to time. And if you're in the tech industry or you're just curious about what's going on in the world of free software and open source,

don't miss an episode of Linux Action News, linuxactionnews.com. Christian worked with us to make sure that we had the most accurate information possible. So literally the moment the NVIDIA news dropped, Linux Action News dropped. So that way you could have the context and the information first and you could make sure that you were commenting or if you were saying something online, it was accurate, right? popping off with our opinion, but JB listeners got the facts just by being subscribed to Linux Action News. So it's something to consider because when something goes down, we're going to try to cover it there as accurately and as fast as possible. So linuxactionnews.com. Big plug for that, I guess. Hope to see you next week. Join us. We'll be live on the Sunday, but of course,

you is always, always welcome to subscribe, enjoy the show any way you want. Live is just one of the options. And of course, we appreciate any value you can return to the show. Maybe it's a membership at Linux Unplugged Core. Maybe it's jupiter.party for the whole network. Maybe it's a boost, or maybe it's just telling someone about the show. That matters a lot too. Maybe it's a review. I don't know. I don't know you. I can't tell you what. But we appreciate you listening. Thanks so much for joining us on this

week's episode, and we'll see you right back here next week. Thank you. so obviously to make it clear because i was not sure that no nvidia driver for the future is only NVIDIA driver for the future is only possible with the right current version of the NVIDIA GPUs and the future GPU cards. And it is not possible to use that one with the 1060, for example, right? No, not with 1060. Wow, is the 1060 that old? Wow. Yeah, it looks like. Yeah, but in fact, so it could be that if for the 1060 a working road driver is running,

then somehow we could profit from that new initiative, right? Maybe. Yeah, maybe. I mean, I'm not sure, because I mean, once again, right, the driver consists of multiple parts, right? This is a current driver, and then there's a user space with Mesa, and then there's also the firmware, and we need all three of them to be able to work together, and I don't think the firmware can work with such old cards. The other thing I had just because it comes in mind now,

the firmware, if I understood it well, before, all right, the firmware actually has some shared code that is also used on Windows, right? So the Linux and Windows driver with the firmware share some code, right? I usually get someone from NVIDIA to go in detail. I mean, I know their driver in general is shared between Windows and Linux,

and if there is exactly the same firmware, I think it is, but I don't know, between Windows and Linux systems. I mean, they try to, I mean, of course, for ease of maintenance to keep things as aligned as possible between the two. Here's what I'm wondering, and this is, we'll just see,

and it's going to depend on developer resources, but imagine we're four or five years into this, right? So we've all been using an open source NVIDIA driver for a while. We've got the open source AMD driver. We've got the open source Intel driver. Do we start to see one common interface to start to change some of these settings or whatever is maybe gets exposed that in the past required an amd control panel and an nvidia control panel like

do we start to see maybe some some basic settings just get baked into the plasma desktop or gnome desktop and on the back end they're figuring out if that's for an NVIDIA card or an AMD card. It feels like the integration and control that could eventually come to the desktop is going to be at a whole new level when all of the major video card drivers are open source. Yeah, I think my thoughts around there is that they are continuously expanding what Vulkan can do. So, I mean, I mentioned this to you afterwards over interview, but I'm really excited about the fact that there is Vulkan extensions coming out to do H.264, H.265, and VC1 encoding and decoding now.

Which means that instead of, up to this point, having to deal with a special encoding library for NVIDIA, a special encoding library for Intel, and so on, we can now, as application writers, right, have one API we target, and we will get GPU accelerated encoding and decoding. So this is something I'm really excited about. And I think, you know,

I think we'll see a lot more things moving into Vulkan over time, because I think they're definitely planning to go on beyond, as we see with the video, just doing 3D graphics. Christian, as a last question, I'm curious about what was your personal favorite part about working on this initiative for you personally? Well, the main thing, of course, is that I know that for Fedora users and RHEL users too, of course,

but it's been a pain point for years to have to then deal with going out or first realizing, oh, okay, I need a driver for my hardware because I think on Linux, people sort of assume everything is just there. And then maybe not finding it or maybe they found the wrong version

or they found the version that was harder to set up. I mean, I know a lot of people, of course, ended up getting the upstream NVIDIA driver, which they then have to build themselves. And of course, I did that myself back in the day, and it was a great learning experience. But it would have been a lot easier for people to, for instance, have found the repository of RPM Fusion to just get it pre-set up for them very easily. So I think just being able to get beyond that and sort of say, yep, graphics is not going to be something where I, you know, I have to read a review for a new photo release and people saying, Oh, this broke. That's the big thing for me.

Oh, I can't wait.