LINUX Unplugged - 525: Beating Apple to the Sauce

Episode Date: August 28, 2023

We daily drive Asahi Linux on a MacBook, chat about how the team beat Apple to a major GPU milestone, and an easy way to self-host open-source ChatGPT alternatives. Special Guest: Neal Gompa.

Hector Martin is asking the Asahi Linux community how they feel about adding some what he calls, quote, trivial telemetry to the Asahi installer. And he says, right now, we have no idea how many installs we have. I can guess based on installer downloads, but it's not accurate. He'd like to have the installer version, device type, model, macOS, and firmware versions, and what image you selected, and what OS the firmware and the installer selected. And he says, no serial numbers or anything like that. Maybe partition sizes rounded to the nearest 10 gig. How do we feel about this? This seems like, to me, pretty understandable, but telemetry is telemetry.

Yeah, I think I'm at the point where if that's what the project needs to sort of just keep this momentum going, have at it. Plus, you're going to already have to buy a Mac to get down this path, so... You're already going to have telemetry collected. Seems like a reasonable compromise. Totally unrelated. I just love in this post at the bottom, Hector notes because he's a total nerd.

For the privacy conscious, I estimate that info to have about 16 bits of entropy, give or take. Hello, friends, and welcome back to your weekly Linux talk show. My name is Chris. My name is Wes. And my name is Brent. Hello, gentlemen. Well, coming up on the show today, there have been some significant developments in the effort to run Linux on the Apple M hardware. So we kick the tires and we'll answer the question, could you buy a Mac to run Linux yet? Are we there yet? Plus, our pick this

week is a barn burner. You can run the latest open source chat gpt alternatives locally fully self-hosted and it's super easy we'll tell you how coming up in the show and we'll round it out of course with some feedback some boosts and a lot more so let's say good morning to our friends at tailscale oh hi hi tailscale tailscale is a mesh vpn protected by wire guard go to tailscale.com slash unplugged and you can try it for free for up to 100 devices unlimited subnets and support the show We love it

We use it on the back end, we use it for our personal stuff It's really pretty fantastic tailscale.com slash linux unplugged And of course, time appropriate greetings to our mumble room Hello virtual lug Hello Greetings Nice little showing time-appropriate greetings to our mumble room. Hello, virtual lug. Hello. Hey, Wes. Hey, Wes.

And hello, Brent. Hello. Greetings. Nice little showing in there today. It's great to have you guys. We'll be getting into it today, I assume. So happy to have you there. And a little bit smaller quiet listening lobby today,

but still some folks up there. So hello up there to the quiet listening. A reminder that it's a great option to listen to the show live. Or go hang out with Brent in Berlin Friday, September 8th. It's coming up pretty soon. Get your bags packed, Brent. Yeah, they're just permanently backed at this point. Just easier.

Yeah, I'll be back for the next Cloud Conference which is happening the weekend after the 16th and 17th. If people want to join that, there's a bunch of really cool talks happening. And you can just join remote. You can be anywhere. So if you're near Berlin, come join us in person. It's free. Uh, but if you're not anywhere near and join us, um, virtually, we'd love that, but really come on the eighth and join us in person at the next cloud office for the little JB meetup we're doing there. This is what left the fourth one that we do in Berlin, which seems kind

of crazy. Had you asked us, I don't know, six months ago, if this would have been the case, I would have said, no, you're crazy, but here we are. So come join me. And I'd love to meet some new people. I know someone just booked a plane ticket this morning to do the meetup. So I'm impressed already. So we'd love to see some folks, some new faces and some old ones too. And I have decided to pull the plug on the spokane meetup uh i think it i think it's just too close to linux fest i assume or i waited too long to have the meetup but we only had one person sign up i'm not sure if there's other folks that were thinking about it but only one side but i thought yeah let's just

let's just end it now and i can cancel my plans and stop, you know, doing all that and just, we'll try it another time. We sure will, right? Spokane, we're coming for you. Sure, maybe springtime. Once it, you know, warms up over there a little bit. Maybe we'll come over there. We'll find out.

But keep an eye on meetup.com slash jupiterbroadcasting for future meetups. And of course, Linux Fest is coming up really soon. Well, there's been some pretty big news over in the Asahi-verse. Fedora and Asahi have teamed up, and something called the Fedora-Asahi Remix will soon be a new flagship distribution, providing a polished Linux experience on that Apple Silicon. I've been running this for weeks before they even kind of made this official announcement Linux experience on that Apple Silicon. from mac os and uh neil's joining us in the mumble room today and he's working with the remix to try to bring essentially fedora in line with what the needs are of the asahi project to make it all run

and that's why they're doing this in its own remix for now uh and neil i saw your flock presentation which i'll put a link to in the show notes where you and david introduced some of this but i guess could you give us a quick recent recap of why you guys created a separate remix instead of just somehow enabling all this upstream in Fedora? I think the misunderstanding here is that we're not, the thought that we're not enabling this upstream in Fedora. We absolutely are. But part of the rules for how Fedora variants are classified and marketed, things that are mainline Fedora Linux need to all be built from components that are

shipped within solely the official Fedora repositories. Now, almost everything around the Asahi integration work is in the Fedora repositories except for the kernel u-boot and mesa those three components are still out of tree not everything's been merged into their upstream uh projects right fedora is a very upstream proof centric model and so because they're not all fully merged upstream the stuff that still remains you know separate we ship in a copper uh for the purpose of integrating into a remix and so then we take all the fedora content we take the copper we put those together and produce um images that are then made available um as the fedora saiy. Now, because we're doing that, we have to tweak the name a little bit and all that other fun stuff, hence Fedora Sahi Remix.

So we are actually aiming for integrating all this stuff in Fedora. All the user space fixes, we're not forking most Fedora packages. They're getting fixed in mainline Fedora. So things like getting WebKit and Chromium and all these other things to work properly with 16K pages, dealing with bugs and crashes and other stuff related to just running desktop applications on ARM and assumptions being invalidated, those are all being fixed within mainline Fedora. We're doing this in such a way where it benefits anybody

running a Fedora desktop on an ARM platform. Yeah, and so inevitably, I mean, hopefully, all of this will just sort of be collapsed into Fedora Workstation. But in your Flock presentation that we'll have linked, your cohort, David, there also gave a pretty good explanation of why this is a good effort overall for Fedora. And I'll play that clip. Back in 2020 and even before, one of the things that Neil and I worked on was getting ButterFS in Fedora. And ultimately, this led to a change for Fedora 33 to make ButterFS the default file system.

And one of the things we had to do as part of this was, of course, making sure that ButterFS would behave well on all architectures supported by Fedora. And one of the architectures supported by Fedora is AR64. And one of the things we quickly discovered back then was that there aren't that many platforms available in the market that are widely available,

are reasonably cheap, and can be run 24-7 without self-destroying after a little bit. So we were running things like file system stress tests, and if you run file system stress tests on Raspberry Pis, your Raspberry Pi will probably die after a month-ish or so of operation. They're just not designed for this kind of workload.

And while you can definitely get server-grade ARM64 hardware, you probably don't want server-grade ARM64 hardware in your apartment, both for power usage and for the noise, and even more so during the pandemic. That was definitely not something you wanted to deal with. So we were keenly interested in having a usable ARM64 platform that would be well-supported for this kind of workload. So it sounds beneficial for the Fedora project, but I also kind of get the sense that it's beneficial for asahi too because there is a structure to work here with there are interested parties that are working upstream with stakeholders in fedora and i saw this post on mastodon by hector he wrote

that he's he's regretting that they started with arch it sounds like um he writes i'm going to be honest with everyone i'm getting really tired of Arch Linux ARM. Missing packages from upstream Arch that do build properly out of the box, random broken package builds, broken dependencies for years on end, missing rebuilds after ABI bumps of dependencies, and now Firefox fails to build with WebRTC, so let's just disable WebRTC. And the maintainers are generally unresponsive. I apologize to all Asahi Linux users. You deserve better. When I chose Arch Linux ARM as a base, I didn't realize it would have so many basic QA issues. We're working on better options. Please just give us a bit more time. This sounds like before the Fedora news was public, because this was back in

March. And I feel like that gives us some of the insights into the incentives on the Asahi side, where they might want to interface with the Fedora project as well. And it feels sort of naturally in line with Fedora's focus. You know, I approached Hector and the Asahi team, I think all the way, I think it was like initially a month or so right after the asahi linux project launched on christmas 2020 you know i'd followed them very closely and talked to them like i think the only reason that maybe that wouldn't have been possible to start with fedora asahi remix uh out of the gate was that i didn't have a way to do anything like as a as a, I didn't have access to hardware or the ability to have a good feedback loop to make sure that things were in place to work. Those things got hammered out over

the course of a couple of years. And like, I think since the beginning of this year, we've been doing everything in parallel for Arch and Fedora in the background. we actually started, I think sometime last summer and got quickly in sync with, with the arch experience and then, uh, got better than the arch experience, uh, you know, from, from our perspective, uh, within a, within a month or so after that. And then we just, we maintained a solid parody for, uh, for a long enough time that we started getting comfortable with digging in past the hardware integration parts and starting to deal with the desktop experience parts. And that's what a lot of this past eight, nine months has actually been about. Seems like that's kind of one of the areas, right? Like the integration, the desktop experience, that focus of Fedora is one of the things that maybe we're getting now that we weren't quite getting in the same way in the Arch universe.

Rounding out maybe the rough edges on Plasma to make it work correctly on ARM with the acceleration and things like that. Because I've noticed that's been steadily improving as I've been running it over the weeks. There used to be little issues that I'll talk about that have cleared up. So that's been nice to see steady improvement there. In fact, that's one of the other big announcements so we have this official fedora si remix that will be it's available now kind of you know for testing but it will be officially launched very soon they were initially targeting the end of august you know about now ish but there's just a few more things that need to be

worked out so that's going to probably just be a few more weeks. On the plus side, though, they don't expect any more breaking changes between now and the actual release. So if you do install it now, you should be able to just do a DNF update and get onto the official release. Yeah, which I plan to do. The other thing that's really awesome to see, and they beat Apple to the punch here, which I think is fantastic,

And they beat Apple to the punch here, which I think is fantastic. Their M1, M2 3D driver is now OpenGL 3.1 ES conformant. And that's really great for a lot of reasons, but it's not even something Apple's managed to do with their own hardware yet. And that's why I think it's pretty great pretty great it is they say quote the first conformant implementation of any graphic standard for the m1 and we don't plan to stop there so we've got just kind of recently in the last few months we've gotten open gl es 3.1 vulcan is a work in progress and they're proceeding there and it just is it's the development sort of leapfrogs like they had open gl 2.1 and they just went right over 3.0 and went to 3.1 and so they're just making really incredible progress

on the video front and i've noticed a difference it seemed like at first you know there was like all the initial progress and we were excited and then kind of back into the skunkworks times and things were happening but not as many um exciting announcements it's it's, it's so nice to get these little updates that just sort of confirm like, yeah, I mean, there's a lot of work that is still happening, but things just keep getting crossed off the list. Linode.com slash unplugged. Head on over there to get $100 in 60-day credit,

and it's a great way to support the show. A lot of you have gone over there to spin up an IPFS podcasting note. Well, okay, about a dozen, and it's a great way to support the show. A lot of you have gone over there to spin up an IPFS podcasting node. Well, okay, about a dozen, but that's still pretty great. And, you know, Linode is leveling up right now. All the developer-friendly tools that we've always known, like the Cloud Manager, the API, the command line tool, all the things you need to build and scale in the cloud,

that's still there, but now it's all combined with Akamai's power and global reach. And they're investing in Linode services to offer more cloud computing resources and tooling while making sure it's still reliable, affordable, and scalable for yourself or a business of any size. Akamai is the king network out there, and now they're working to make Linode even more powerful. And as part of Akamai's global network of offerings, data centers are expanding worldwide. They're going to give you more resources, more places to spin up a data center to help you serve your project, your community, your business, or your friends.

So why wait? Go see why we spin everything up on Linode, why our audience spins things up on Linode, and why they get that $100 credit and really kick the tires. Go experience the power of Linode. Now Akamai. Go to linode.com slash unplugged. To power of Linode. Now Akamai. Go to linode.com slash unplugged to learn how Linode, now Akamai,

will help you scale your applications from the cloud to the edge and see why we've chosen to use it now for years. It's really the one I feel comfortable running my business infrastructure on and delivering all of the show assets to you. And there's a reason why we choose it. Linode.com slash unplugged.

Well, Chris, you mentioned you'd been running Fedora Asahi Remix. You have something of an experience report for us? Yeah, I've been kicking the tires. And you've got to go in with some expectations set to medium. And the project has a wiki on their GitHubithub that is a what i'd say is a feature matrix you can look at the model of mac that you're considering in the m series from the mini m1 all the way up to the m2 ultra and you can get a good idea of what kernel version you're going to need to have your thing supported what's a work in progress what doesn't work yet so i want to link

that and mention that first because you're going to go into this and with some limitations um things like uh video decoding hardware video decoding that's not there yet right that's just not ready yet um video out from my macbook it's not there yet speakers are still disabled because the damn speakers can blow themselves up there's no firmware limiter so until we really have that solved best safer to just turn them off not have sound okay so if you're going to use it with any you know serious amount of time you'll probably want to plug in like a usb audio thing which i have i like it so you're using the built-in display but you got to get your external sound yes i mean at least external audio devices usb ones even are super

affordable and pretty much available everywhere. Yeah, cheap and easy. And that stuff has worked great with Linux for a long time. And at the end of the day, that is what your requirement is. Does it work with Linux? And then the sub-requirement of that is, does it work with ARM? And if it's an open source and it's built into the kernel, the answer is generally yes. And these audio codecs have worked for years.

Performance has always been good it's always even early on it's like oh man even when it was cpu rendered it was still pretty good uh but recent improvements to the gpu stack are very noticeable in plasma because i'm doing a scaled up version of plasma and that always exposes a little yeah of course yeah whenever you're doing the pixel doubling thing. And recently-ish, in my experience, because I'm only jumping in there a few days a week, recently-ish, I had my graphics acceleration completely break on me. And I thought, okay, this is probably something to do with my recent update.

So I went to the new discourse that's now on the Fedora discourse. And right there, boom, I saw that there is this transition from a 4k kernel to a 16k kernel and maybe neil you might want to explain that a little bit and in that transition i needed to make a few tweaks remove some stuff create a file install some stuff and it's all really simple all in there i'll link to that and so there has been every now and then like little fixes you have to do to just kind of keep things running. But on the order of like how hard it is, it's take five minutes, copy the commands and do it. So it's not very hard. The way Apple Silicon itself works, state of the world today, Apple Silicon, in order to work correctly with all the hardware enablement and stuff like that, Linux has to be compiled with 16K pages in its memory management subsystem.

This is an important foundation that affects literally every aspect of the Linux kernel. So historically for the Fedora Sahi Remix, the main kernel package called just kernel for Fedora Sahi Remix was modified to use 16k pages. As part of some upstreaming work, again, as I mentioned before, we're trying to bring all this stuff as close to Fedora as possible and getting things upstreamed as aggressively as possible. We realigned our kernel packaging with the mainline Fedora kernel package. And so the regular kernel package goes back to 4k pages,

which is the default in Fedora and is what is required to support almost every other ARM platform. But we still need 16k pages for Apple Silicon. So there's now a new kernel flavor called kernel 16k. And so that request is to transition everything over. We give people a transition period, but you just got to switch everything over yourself in this timeframe. New installs are obviously not impacted by this because we've already made these changes for the image builds. need to do this before we go to the next kernel rebase where the old kernel variant, the regular kernel package will go back to being basically like the Fedora kernel and will no longer successfully boot your system. I think it's kind of an example of these are some of the things you have to ride, but these are getting less and less and they're getting sorted out right now. Yeah, that's the reason why we didn't just say,

this is GA now at Flock, was because we have a to-do list, an internal one, of all the things that we need to be release ready. And we needed to make sure that certain major backwards incompatible changes needed to be done now. Now, that's not to say that there won't be more in the future because something will force us to,

but we don't expect any. I think the question that I wanted to try to answer this episode is, could you buy an M-based Mac today with a reasonable expectation of running Linux on it? So you wanted some hardware that you could dual boot. You're thinking about getting this for the battery life or whatever it might be. you sad could you at least have a satisfactory experience running linux and i think that answer if you're okay with the exceptions in that feature matrix is yes i hooked up a usbc dock and had my audio interface plugged into that and that worked for me and it it was really well performant the battery life is actually pretty great uh you know it's hard to really nail it down because it really depends on your screen brightness and all those other factors but you know hours six seven hours of battery life isn't uncommon for me oh wow that's pretty good that's pretty darn good yeah like um in general um i

would probably say the older the apple silicon mac generation is, the better you're going to be at this. But, you know, over time, you know, we're going to be in a position where even uplifting to new Apple Silicon platforms should not be so difficult. Part of this issue right now is that it's not just Apple being fairly aggressive about reworking a few things here and there when they're trying to expand and fill out the feature set. I think Hector made a Mastodon post about how the Mac Pro was clearly not the architecture that Apple wanted, and it was the architecture that they settled with to get the product out the door. I agree, yeah. As we, you know, fill out the hardware support matrix and get these things in place and get the drivers mainlined and whatever, going forward, I kind of expect to mostly be telling Linux that these new hardware platforms exist and then pointing them to the right drivers. Yeah, and honestly, Neil, the end user experience, while you guys are on the front line, the end user experience is I just do DNF update and I just kind of get the new stuff and stuff

generally just works better every now and then I've had to fix something. And that's, that's, that is, if you're okay with that experience, I think it's ready. And you know, if this is a hardware platform you want to buy into, I mean, there's a lot of good choices, but if this is you've made up your mind, I think you'll be pretty impressed with how far it's come. You are on a limited set of software availability because it's ARM compared to x86. And there are some things that are just less tested that's getting better as there's going to be this Fedora remix. It'll get more users. It'll get more testing. The web browser will be your friend. Apps that maybe you installed as standalone apps, you're going to have to probably fall back on their web versions, but it's not all

bad. Sometimes it's actually faster. That's my version of using it on the desktop is you could, and it's going to work pretty good, especially if you just live on the laptop and you're not trying to do a bunch of external screens. Makes me wonder, can you use some of the old tricks from the more common dual booting days? Like maybe you're on the road. Could you boot up the Asahi partition as a virtual machine on the Mac side if you kind of need to rely on that while you're mobile? And then when you're home, you like boot back into Linux and plug it all in. Run full time again.

Yeah. I would actually recommend you use mainline Fedora ARM for a VM because the way Apple configures VMs on Mac OS is that they are actually essentially emulating standard 4K ARM UEFI based platforms. In fact, I think the developer documentation for, you know, if you wanted to write a program to boot up a Linux VM actually uses Fedora Linux as its example. I think it's right. And in the screencast, too, and they're showing the Linux VM actually uses Fedora Linux as its example. I think it's right. And in the screencast, too, and they're showing the UI, it's Fedora. First ones, I think they use Debian. And I think they, for some reason, they switched to Fedora later on. Yeah. Just use Fedora KDE on ARM in a VM and, you know, you'll be fine.

Here's where I think it's a slam dunk. You can now pick up an M1 Mac mini a lower end one for under 500 and you know that's going to be even cheaper when the next one comes out and if you are running this thing headless this could really be a sweet home server and so you know these are the kinds of things where maybe you're not buying one today but when you're looking at your next home lab and you maybe you want to go lower power and you want to go silent the mac mini m1 idles at 6.8 watts and when you ramp it up and it's doing things that put a raspberry pi or an odry to shame it's only using 39 watts of power at full kilt and it runs absolutely silent and it's also probably the best supported

device for asahi linux and so these things are going to be coming up for sale especially because mark my words the m1 platform in a couple of years will be abandoned by apple they have done this before when they move from the 6080 to power pc and when they move from the power pc to intel the very first machines they released after those transitions received the most limited amount of support by apple because they were the most technologically naive and they wanted to move on and i believe it's probably going to be the same for the m1 mac mini and so i feel like it's going to be a perfect one for the linux community to embrace now i i think since this thing it's all soldered on there, you'd probably want to get it with a fair amount of RAM and a fair amount of

storage. But, in a year or so, they may have Thunderbolt 4 fully up and working. Ooh, that'd be nice. Then you could attach PCI storage. That'd be plenty quick and low power. And so I think it is there right now if you want to build a home server. And I think

we ought to take a Mac Mini that we have kicking around. I'm not using it at the moment because I was afraid it suffered water damage. And we convert it into a Nix Bitcoin node. A silent, high performance. It's got a two terabyte NVMe in it. It could just sit there and hum along all day long, sipping six watts of power. I like the way you think. I think it could be sit there and hum along all day long, sipping six Watts of power. I like the way you think.

I think it could be good for that. Now, I don't think it's great for use cases where you want video accelerated decoding for like Plex or jellyfin. And if you need a bunch of external storage, you don't want to use USB. It's not going to be ideal for that. But if you're looking for something that just has good on device compute and good fast storage,

I think it makes a great little home server, potentially, depending on the task. I also want to point out something really fun that I think a lot of people missed. It was briefly mentioned by Davida during our blog presentation, I think, but it was also mentioned by Joseph Basik, one of the Butterfest developers on Mastodon and on his blog, that some of the upstream ButterFS CI now runs on Fedora Linux for x86, but more importantly, Fedora Sahi Remix on ARM. So we are now regularly qualifying upstream ButterFS code on this platform as part of just doing development. And this was actually one of the, as Davida alluded to at the very beginning of the story that you clipped in earlier, that this was the motivator. Because I have in my house a smorgasbord of dead ARM devices from doing this.

Because I have burned out every eMMC you could think of and way too many SD cards. These things are way sturdier and way hardier and can survive being run as test harnesses. They're really the first kind of general purpose you can go to a store and buy an ARM workstation. True. It's kind of wild that nobody else has really done this. There has been attempts

and there's different types of hardware out there, but nothing at this scale are like this. And in a weird way, seemingly impossible against all odds truly the impossible mission it works in a lot of ways better than most of like the ppc the power pc linux ports did and in a lot of ways we're already further along than some of the previous attempts in history to get linux working on mac hardware it it's exciting it it's reminding some of the previous attempts in history to get Linux working on Mac hardware. It's exciting. It's reminding me of the golden years period of the Intel era. I remember when Matthew Garrett was doing a lot of good work, and Fedora really was a great OS to run on some

of those Macs. We're getting back there. Collide.com slash unplugged. Attention, security, and IT professionals with Okta. Have you noticed a recurring pattern in the recent data breaches and hacks? Maybe ones you've experienced or ones you've seen publicized? It's often employees' machines or roles or credentials that play a role. Unpatched software, accidental data handling, compromised credentials through phishing. Those are all pretty familiar culprits these days. And the real issue isn't the individual users, but it's inadequate preventative measures. Well, that's where Collide comes in. The solution to change this. For those of you

in security IT that are working with Okta, Collide ensures only secure devices can access your cloud apps. So say goodbye to compromised credentials and the hassles of managing diverse operating systems. It's all in one dashboard, even the Linux desktop. Plus, you can empower your employees to fix their own devices directly without burdening IT through some clear messaging. You should go see this firsthand. Go discover Collide at collide.com slash unplugged. They got a demo right there you can watch. You can see how it works, and it's a great way to support the show.

That's K-O-L-I-D-E dot com slash unplugged. We'd like to say thanks for the feedback that was sent in this week. Really appreciate that. If you'd like to leave us some, linuxunplugged.com slash contact for that. Peter sent us a note. Dear Linux Unplugged team, I'm searching for an application where a client runs in the background on each team member's machine and rings when someone initiates a direct call, ideally with screen sharing functionality built in. Not interested in solutions that require sending an invitation link, though, like Jitsi Meet to initiate a call. Do you have any suggestions? Sounds like Peter's kind of looking for a Skype-style solution. Yeah, I was going to say something in that space, VoIP-like products, but a screen sharing, you know, you're probably going to need that to be supported decently, at least on your side, maybe on both sides.

Mm-hmm. We might have to turn this one over to the audience. I bet they have some ideas. I was trying to rack my brain. is. I was trying to rack my brain. You know, I was thinking a SIP-based solution or something, but Jitsi Meet would be useful, but it's web browser-based, so you can't get an incoming call.

And a lot of the SIP solutions don't do screen sharing, as far as I know. There's no setup for that. I mean, you could do something like Matrix or Mattermost, because then you could do calling and screen sharing and meetings in there, and you would get a notification from the app. app but yeah i think we should turn it over to the audience

and see if we can get any suggestions this would be fun for us too the reason why i went sip is because i would love to have a system where we could have both soft phones so you guys could call into the board but also like we could have physical phones at our desk. Like, Brent could have one up there, and I could have one. And we could, like, have extensions, and we could call each other and be like, hey, what's going on? You know, do, like, a group call. Okay, well, this is a fun idea. I know.

We can make it happen. I think we could. I think we could. But anyways, let us know. Boost in with some feedback for Peter, or go to linuxunplugged.com slash contact. Now, Morgan sent us a bit of a story, but I think it's one worthwhile. So sit back and relax while I read this one.

A quick story on how probably not well-known XFS feature saved my butt recently. Working on HPC, I have various clients with petabytes upon petabytes of data. While this specific client has a dozen bare metal boxes with JBODs. In this case, I was reinstalling the OS, which was Rocky 8, if you're wondering, on one server, which along with its own internal RAID card and its 24 3.5 inch bays had another 24 bay JBOD connected to it. Both the internal and the JBOD RAIDs were 110 terabytes, each full of client data. Well, I go to reinstall the OS and the provisioning system, XCAT, ran a wipeFS across all of the disks during its kickstart installation, unbeknownst to me. The server

comes up, all looks good, except now devSDBb and devsdc have exactly nothing. No partition table, nada. I can see the hardware RAID is fine. Virtual disks are still showing their 110TB capacity, but from the Linux side, I just have nothing. I try various things, all to no avail, so I'm about to just remake the broken arrays into fresh XFS partitions when my coworker mentions that XFS actually writes its superblocks to multiple places on disk,

so if one gets corrupted, there are potentially backups to fall back on. And, sure enough, XFS Repair was able to find one of those other superblocks, an entirely fixed one of the VD's partition table and the data was entirely intact. Too bad the other VD was originally a Butterfest file system. And while Butterfest does do similar things, I was, it was simply too far gone. A win, but a loss. XFS saves the day again. Also, Chris, you asked about inlet temperatures for servers in Linux Unplugged recently. My data center runs hot and delivers 81 Fahrenheit air to the servers. I've been there twice, though, when the cooling systems failed, and once it was long enough that we had to start shutting servers down.

During that experience, I found that the average server was starting to throttle once you maintained an inlet temperature of about 90 to 92 Fahrenheit. So I would say it's unhealthy long-term to run commercial-grade servers with inlet temperatures above 85 sustained. Thanks for all the shows. You guys are awesome. Well, thank you for the story and the data on the server temperatures, trying to collect that, kind of massage it all together into some signal. And I think the message I'm taking so far is and it's obvious when you start getting into the high 80s and 90s it's getting harder and harder for the equipment to ambiently cool the cpu and pretty soon the air it's pulling in is getting close to like the temperatures you

you know the cpu should be running at and you can't really cool much but I'm amazed at the shenanigans I've gotten away with, you know, cause there has been days where I forgot to check and I have not set up an alert and I'll go look at the history and I'll be like, Oh, that that's 91 degrees. That's a hot one. Imagine you talking to the server sort of in like a Janeway Voyager style.

Yeah. Yeah. You can do it. You know, you'll get us home. Yeah. Thank you you can do it. You know, you'll get us home. Yeah. Thank you for reaching out, though. We really appreciate it, Morgan.

And sorry to hear about the data loss, but glad you did recover some of it. And thanks for sharing your pain. Yeah, really. I want to remind everybody that Podverse has a $1,200 bounty for complete Android auto support. They have details on their GitHub page. They've got a few other things going on over there, too, too but could be a great opportunity to contribute to a gpl podcast app and get android auto working for everybody i think they got carplay so now they just need to finish it off so close so close check out the pod verse github for deets on that and before we

get to the boost i want to just address a question that we got into a different show but i thought we should talk about it here it's sort of like sat skeptics why why sats to support the show and not you know whatever it could be a different cryptocurrency or could be a different payment system and i get it too because like the last couple of years it has been so crazy watching all these scams and like the monkey jPEGs and all that stuff just going for. And now they're going for like nothing, too, which is hilarious. Yeah, I think like one of the most famous ones is like Jack Dorsey's NFT tweet or something that sold for like some ludicrous sum of money and now is worth like, you know, nothing. And so it's I get it.

In fact, I think it's probably kind of it's's intelligent to be skeptical by default on this kind of stuff. I want to give you a thought experiment to play with for a moment and see if this helps. Imagine if somehow Linux wasn't popular until just about now, or maybe a few years ago in the tech cycle. Somehow the cloud had come and gone and, you know, we'd had that boom and bust cycle somehow maybe windows was huge or whatever and so the industry understood the value of a really solid cloud platform for application development and they saw the potential so they at this point in the industry would ape in like crazy vcs would be d gens and they'd be backing every crazy linux project out there that could become the next standard platform and so every single distribution crap or good would have a million dollars or 10 million dollars or 100 million dollars injected into it by vcd gens and then they launch marketing departments and that's that's predominantly what

they focus on and those marketing departments do everything from pump and dumps to hype on social. And it would totally destroy the credibility of Linux. And you'd have low information Windows users who don't really know about Linux, who'd be kind of ignoring it, thinking, I'm just going to wait till this goes away. This whole thing is just a total crap show. I've heard there's this Ubuntu Satanic Edition. How would they even know the difference between Ubuntu Satanic Edition and Ubuntu Mate? To them, they just think Ubuntu is some sort of scam distribution. They wouldn't understand because they hadn't spent the time that there's Debian's and Fedora's out there that are real gems that provide true value. It would all just look like Hana Montana Linux or Ubuntu Satanic Edition from the outside.

montana linux or ubuntu satanic edition from the outside but somebody who truly understood linux and followed it would know there is real real signal in that noise that's the bitcoin situation the problem and with the problem with bitcoin and it was a real real pain in the ass for companies and vcs is there is no bitcoin ceo there is no company behind bitcoin there is no commercial interest or group you can't buy them out and then own the technology and acquihire them like you can't with a lot of open source things so we had to copy it we had to make coins and all of these copies all these stupid dog coins and all this other crap. All has people behind it, running it, making money when the price goes up and then they dump on people. And they are quasi Ponzi schemes and they are awful.

And all of it comes because Bitcoin proved the idea. But the problem with Bitcoin is that the development is decentralized. The node network is 15,000 plus nodes out there, totally decentralized. And the user software is decentralized. And the whole damn thing's free software. Well, what a pain in the butt for all these crazy VCs that want to make a bunch of money. They couldn't solve that problem, so they created coins. And it destroyed the reputation

of the original creation, which is truly great and is an alternative system outside the commercial finance system we all exist in. Bitcoin is the free open source solution that exists outside the proprietary commercial system

that we all used for our day-to-day transactions. And when you look at it from that perspective, it does seem like an odd thing and it does take a little bit extra work to go use but there is genuine innovation there not so much with all this other crap so i completely understand why people be skeptical but i invite you to experiment with us participate in the value for value exchange using a boost

with some sats and experiment and see where it goes and i think it's going to be a great time to do it because i would expect the price is probably going to remain pretty low maybe until the election it could run low for a while so it's a great time to experiment we're not suggesting that you invest or hodl saying grab a few sats send them our way and we do what we like with them we could sell them we could hodl them we could put them into channel capacity as a a business, we make that decision. And there's no middleman. The entire infrastructure is completely self-hosted. The sats are sent over an open peer-to-peer network and land on my system that I have here in the studio, all using open source. There's no Patreon in the middle that's going to go screw everything up and have a 30% decline rate for three weeks. It all comes directly to me.

If it doesn't work, it's my fault. That is incredibly empowering for independent media. It is a fundamental game change for independent music and maybe one day free software. And I think right now we need this in independent media. We need this in independent media. In all areas, news, entertainment, education, independent media is more important right now than it has ever been in the history of media. And now we have a truly decentralized free platform that uses free money. And you can self-host the entire infrastructure if you want, or you can go pay some service provider if that's your preference.

It's a cool technology, so we invite you to experiment with us and with that said let's get into the boost and now it is time for the boost oh yes it is have we decided this we just pronounced the geek for this one or tech geek i guess yeah with 311 806 sats you are our baller sending from the podcast index coming in uh hot with a follow-up from episode 522 nixos rocks there have been a few learning curves with nixos but it's definitely working for me on my dell 7480 laptop he says chris i'm sorry the only rv hook is if i'm aware of in my area are just the campsites at disney's fort wilderness resort but the family and i frequent there staying either in a tent or in the cabins they do have rv sites as well i've always wondered about that you know you know for

us rvers they have like these all-inclusive like you show up at the resort and then they shuttle you to the amusement park. Okay, that doesn't sound so bad. It could be fun. Sounds really expensive. Indeed. MCZP boosts in with 55,408 sets. Jupiter Party member here, first-time booster. Y'all have really helped me through some tough times.

Boosting is the least I can do. It's another zip code boost. Oh, better bust that map out. Oh, good, you brought it. Good. Right here in my back pocket. Look at you. That thing is well-traveled.

This is, yeah, okay, well, there's a few coffee stains. I know, I know. This is a postal code in Minneapolis, Minnesota. Hello, Minnesota. Yeah, oh, I see. MCZP. I think there's a code right there. Thank you for the support.

Trey Forden boosted in with 44,444 Satoshis. Over and over, you guys keep banging the NixOS drum. I briefly tried it, but couldn't get over how hard VNC server was for me to run properly. However, NixOS has fantastic support for RDP. During my lunch break, I created a VM and tailored my Nix files to my liking. After copying the config off, I deleted the VM and began to set it up again. I get it this time, guys. Immutability is cool. Having a basic config that describes my ideal server environment makes the idea of loss far less scary. Once I figure out how to describe my virtual LAN and OpenSense network for labbing, I may finally feel comfortable switching from Arch and Ansible. That's pretty great. I wonder if we should set a bounty for a challenge to not mention Nix for an entire episode. Yeah, like 400,000 sats and we can't mention Nix. That would be tough.

We have to put that in the boost, the request that we don't. You think we could? Is it like we owe 400,000 sats? It's like a drink challenge sort of thing? It's like me trying to do an episode of self-hosted and not mention Home Assistant. It's just me trying to do an episode of Self Hosted and not mention Home Assistant. It's just pretty much impossible. User 66 came in with 36,000 sats using Fountain, and they got a postal code boost in their way. So get yourself ready.

And also they say, new listener, second time booster. Well, thank you. Okay, this one's a little more complicated because there were two boosts in total. So just looking at the second one, which was the postal code boost, that was for 25,767 sets, which appears to be a postal code in Germany, somewhere near Albersdorf, perhaps? Albersdorf?

Let us know. I don't think we're getting that one right, but... Hello, Germany. That's great. Complete noobs came in with 33,333 Satoshis. Coming in hot with the boost. Greetings, fellow noobs. Because let's face it, we're all noobs at something. Thank you for the kind words last episode about the domain name. I secured it about a year ago, as I thought it would be a great fit for a project I'm trying to work on.

Ooh, that's a little bit of a tease there. Want to know more? When should we check back is the question. Right. We are all complete noobs at something. Tell you what. I feel it when I'm doing mechanical stuff on the car. I just didn't learn any of that stuff as a young lad.

And so I don't even know the names of stuff or what the heck I'm doing. It's just a totally different language. It's scary. It's exciting. Yeah. It's a young lad. And so, like, I don't even know the names of stuff or what the heck I'm doing. It's just a totally different language. It's scary. It's exciting. Yeah. It's a good experience. Torped comes in

with 21,984 sets using Podverse and just says, thanks for the content, Wes, Brad, and Chris. You know, I don't mention Brad enough, but he does a lot

for the show. Yeah, Brad does a lot of behind the scenes. Brad and Brett. They, uh, yeah. They really make it so that way

brent can show up and just put on a smile curious concept boos in with 11,111 sets the big old bag of richards hey i need some advice it's time for a laptop upgrade i'm feeling so conflicted because i want a macbook for its performance and the battery life, but I would really prefer to keep running Arch with Plasma for my daily driver. Should I go for an M2 Pro and just hope Asahi support improves over time? I was also considering a new XPS 15, but I'm unsure if I'll really be able to get 8 plus hours of battery with that one running Linux. Are there other laptops I should be looking at? I really value battery life and fast compile times. Boy. Okay. I'm really glad he added that. Your good choice is go with Lenovo. Lenovo computers are, have a wonderful Linux team that is actively working on supporting models upstream in Linux, which flow because,

you know, the primary effort that they target is going through Fedora, and Fedora requires everything to be upstream. So it goes upstream, and that benefits any distro you want to use, whether it's Arch, Fedora, or anything else that actually, like, follows latest kernels. So that would be the way I would suggest you go. I'm still very happy with my X1 Carbon. I have the Gen 8, and it runs all the distros I want. Fantastic. It's still my go-to grab whenever I just want something quick at home, too, because it's

the right balance of comfortable keyboard, weight, and battery life. And it does a really good job of deep sleeping, so I can leave the X1 kind of just on the table for a couple of days crack it open and it's like an 86 battery we're like with the dev1 if i left it on the table for three days it might actually be like at 10 battery when i open it back up so the x1 is kind of a nice experience in that regard the fingerprint reader just works so when pseudo prompts come up in genome or on the command line i can fingerprint authorize so i i would strongly consider that if you want a reliable linux workstation you get a machine that is linux first if you're going to

do 80 of your work or 60 of your work somewhere in that range in mac os then you could do the dual boot thing i think you'd probably okay with the m2 pro but you're going to be probably using mac os predominantly and then you know over time using Asahi more and more and more as more features come out. Because if this is your main machine, you're probably going to want HDMI out. You're probably going to want full Thunderbolt or USB-C support and stuff like that. Something to consider maybe down the road in like six months.

It might be a different answer. But if you really want one really great solid daily driver, I kind of agree with Neil. I'd look seriously at the lenovo stuff and i'd also i consider the xps not as big of a fan of the more recent xpss because they're kind of going the direction of the older macbooks with the keyboard changes and the trackpad changes the framework 13 might also be a good model yes a good option to choose because because they're qualified and tested to make sure that they work on Fedora Linux. And if it works on Fedora Linux, I know you mentioned Arch. If it works on Fedora, it's probably going to be fine on Arch.

I think the framework stuff doesn't get mentioned as much on this show just because none of us own one, and we don't really have, you know. Not yet. And we need to get one eventually, so that way we can chat more about it. VT52 came in with 8,192 sets. Hey, I'm glad you liked impermanence. I came across this neat hack a while back I thought I'd share. How to unlock a Lux encrypted root file system remotely using SSH and a Tor hidden service with an XOS.

Ooh, this is cool. and a Tor hidden service with an XOS. Ooh, this is cool. So it's like running Tor in the init ramfs, and then you can connect over Tor via using SSH over Tor, and then unlock your drives. But I bet you could do something similar, right, with like if you run in Tailscale or another mesh VPN right in your init ramfs as well.

Hmm. That's something we might have to play with. That could be really nice for just a little extra security on a remote box. And it6 comes in with 5,000 sats using the podcast index. Regarding your short mention of OpenStreetMap contributions, check out StreetComplete. It's up on GitHub or it's in AfterRide. It gamifies adding info and submitting changes in your area. And it certainly got me to go for more walks.

I don't know, Brent, if you've tried StreetComplete. I use StreetComplete. Maybe you might have told me about it. I tried it a couple of years ago when I was first getting, well, I say a couple now it's probably five years ago, to be honest. When I was first getting into open street maps and trying to adopt it as my main solution. And it was fun. I got to say, and as someone who's not like a open mapping expert by any means not even close it was nice to have that like feeling of contributing to an open source ish feeling project without necessarily needing the deep technical knowledge so uh i you know it's been a couple years since i've checked this out but i'm you know we got this boost in and I saw it come in and I instantly installed street complete again. So I think I'll be giving it a shot for the next

couple of weeks and I'll even, I'll try it in Berlin and see, uh, I don't know. That's what I do when I go on trips. I don't think about, I don't even think about it at home. I mean, I'm in the middle of nowhere, so I don't know if anybody even uses maps. Just make it really accurate for yourself, Brent. Yeah. They're still on paper maps out there but yeah i like it when i'm on trips i never think of using it when i'm at home it does make me think i'd really really still love a total replacement for ways like magic earth has the reporting elements in it for great navigation. I really like the way Magic Earth displays navigation.

But in the area where Wes and I live, there is complicated traffic frequently that can just show up, and there's also a lot of cops and a lot of cars that are just pulled off to the side of the road. It is an area where Waze is actually used pretty heavily. And so the network effect is strong with Waze. I mean, you know, I'd

love to try something else, but there has to be users. Yeah, if you want the best info, you go to the app that has the most users who are actually using it to report that info. We need something to scrape the Waze API and then get that into, like, Magic Earth and OpenStreetMaps. Hmm. Do I hear another bounty? I don't know about that. I think it's me. Okay. Hal was right, Boosin, with 2,100 sets.

Impermanence, along with Home Manager, is how I manage my laptop and servers. I use a ZFS snapshot to revert. I'm sometimes not completely sure if I'm keeping around everything I need, but if things work across reboots, then I don't really worry about it. Sounds about right. Yeah. Yeah. Thank you, Hal.

Good to hear from you. Gene Bean boosted in with a row of ducks. Here's a little boost to help out to compensate for a DM I'm about to send. Keep up the good work. Gene Bean gets it. You know what? I happily read that DM. Thank you, Gene Bean.

The Golden Dragon, the show mascot, also comes in with a row of ducks. Alrighty, gang. Last boost for a minute. I got a stack savings for LinuxFest Northwest. If you have any recommendations for me to bring, shoot me an email, and I will start on projects. I have a lot of LinuxFest things on my mind. Also, so sweet of you, Dragon. I think we're going to need coolers. I think we're going to need coolers i think we're gonna need coolers we're gonna need a way to feed i'm planning for 600 i think they'll be i'm betting so i'm thinking if linux fest at its peak had 5 000 and we've been off for a couple years we'll probably have

around 2 000 and probably about 600 of those will want to be fed hopefully the rest just starve it's fine so i'm thinking we keep it simple burgers dogs pizza and then you know we could bring other things too for like brand but in terms of the you know the mass production yeah um but then you gotta you know you gotta keep that stuff cool you gotta you gotta trash that stuff we're gonna have to have a crew we're gonna have to have a crew golden dragon so we'll be enlisting you no doubt about. Wolfman 2G1 boosts in 10,000 cents. Last I heard, Google and Facebook use evaporative cooling in their data centers and maintain 80-plus degrees. My first big tech job was at a streaming video startup, and we ran our data center at 87 degrees.

You know, that's got to save money. If you're not cranking the system is hard right trying to find the right balance of how warm can i be but right it's just funny because back in my day the data centers that i always worked at were like 79 69 68 degrees chilly chilly which was kind of nice in the summer but definitely no 87 degrees that sounds honestly kind of brutal to work in for hours at a time. If you're in there, that sounds kind of miserable. Craftnix came in with two boosts totaling 4,444. I think that's two rows of ducks. Double ducks, lucky double ducks. Lucky double ducks. Is there any reason to use LVM at all? I generally just partition disks

as butterFS for root disks or ZFS for storage pools. What would be the advantage of adding LVM? Does it have any performance impacts by adding another layer of indirection between the file system and disks? I use Lux on every single disk. So is there extra complexity of LVM plus Lux plus ButterFS or ZFS? Is it worth it? No.

Yeah. I feel like the reason you use LVM is you're comfortable with it and or, you know, you're not using something like ButterFS or ZFS? Is it worth it? No. Yeah, I feel like the reason you use LVM is you're comfortable with it and or you're not using something like ButterFS and ZFS. Yeah, that's just it. I have very successfully used LVM with XFS for a very long time on my workstation upstairs and I blow away the root and home partitions

and then every install, I just reconnect to that LVM group of disks. And everything just works because LVM has been around for a hundred years. And I move from distro to distro flawlessly with that. And I don't have to have any complicated ZFS setup or I don't have to have any complicated ZFS setup. It imports really easily.

Installers support it well. Yeah. So there is some advantages there, but if I were to read redo that whole set of discs which is like three or four in there i would probably just do with butterfs today right and also the the the main advantage of lvm um is if you're on a simple file system as i'm going to call it i can can't really, you can't see the quotes, but they're there. Simple file system like X4, XFS, and whatever, you're essentially adding a layer of indirection to allow multi-disk support and things like that. But if you're

using ButterFS, there's not really a ton of advantage to using LVM. And maybe people don't know this, but you can use Lux without LVM. using LVM. And maybe people don't know this, but you can use Lux without LVM. Krafen has asked a question in there. Does it have any performance impact? So I'm curious in the hypothetical case where we're using this particular strategy, is there a downside? Yeah, there is. There's a huge downside. You take a pretty big performance hit because you're passing through all the layers and each layer has to do its own IO processing. And especially when you're doing encryption and the encryption is not part of the file system layer. Now, this is also true if you're doing like, say, XFS LVM Lux, right? You're taking a performance hit. Well, Kraftnix continued here in the second boost saying, hey, organic maps is

also a great iOS and Android maps application. Good UX, good UI, and is just simpler than OSM and. I know, Chris, you tried this a while ago. I tried it when we were in Denver years ago for our little road trip. And I tried it again recently. And I got to say, I was kind of disappointed. I used it just in town here. You know, I'm in a tiny little town. I don't really need it to get around. But I thought, hey, I'm going to try it just to go pick up our little CSA veggie box thing. Cause I, and it kept getting like turned around and lost. And I was really surprised. I've been an OSM and user for quite a while. And I was sort of hoping organic maps had that sorted out when we were in Denver many years ago, was that three years ago? Uh, it was kind of the same case. It like threw me off the freeway a couple of times.

Not literally, but, you know, wanted me to exit off the ramp and get right back on and didn't make any sense. And I was a little sad to see that those directions were, I don't know, still a little bad considering I was doing such simple routing here in my local town. Yeah. Yeah. That's kind of my experience. All right. Well, we're rounding it out. Respect should be earned.

Manor B comes in with 5,600 sats using Fountain. I wanted to thank you for some wonderful shows. Also, thank you for talking about OpenSUSE, a tumbleweed on your show. I find that it does not get enough love from the community, but this show, with Brent using it, does talk about it off and on. Having been using it for the last four to five years it just works ancient machines and new machines intel mac machines it

just works sometimes it needs a little setup i just stopped distro hopping uh he says by the way if you append 38 to my boost you get my zip code six digit zip code i'm to have to open this one up a bit larger, I think. Yeah. Aha! This is a postal code in Bengaluru, India. That's fantastic. That is fantastic. Hello. Hello, Bengaluru, India.

It's good to hear from you. I'm also really happy to hear some love for OpenSUSA. It needs more of it. Yeah. I know when you're on here, Neil, you love getting the OpenSUSA love. It's gross, but it's fine. We have to remind you to talk about Fedora. You're the one I have to encourage to say nice things about open Sousa.

Oh, I love it. It's great. That's been my experience too, is the setups a little, maybe less familiar is how I would describe it. So it took me a little bit more, but then after that, just kind of, yeah, it's great. Yeah. Yeah. And, and, you know, the one thing we hear over and over again is once you get it it just runs for years so ben the tech guy boosts in with 8 865 cents i think i know why the one plus six is one of the

only supported nixos phones it uses the qualcomm snapdragon 8, which has been mainlined so that you can essentially run a generic ARM64 kernel on it. The OnePlus 6, along with the other SDM845 phones, it's one of the best supported mobile Linux devices. That checks out. That makes a lot of sense.

Wow. And Alex just gave this thing to me like it was nothing. Jeez, he's missing out. P.S. If you multiply the boost amount by 5, you'll get my university zip code. Ah, gotta do a little math this time. Come on.

Now I like it. Yeah, it is great. Okay, 8,865 multiplied by 5. 44,325. Right. Which would appear to be a postal code in Akron, Ohio. Hello, Ohio! Nice to hear from you, Ben. Thanks for boosting in, and thank you for pointing out that Snapdragon 845 has been upstream.

That just makes a ton of sense, and it makes me feel like the OnePlus 6 is a keeper, Brent. Yeah, now I don't know if I should stop using it and cherish it forever, or if I should just continue using it, but with something more experimental. So, send your ideas in. Listener Jeff boosted in with 10,001 Satoshis across two boosts. Hey, about touchscreen input, a little hack I've used before was to use my phone's keyboard via KDE Connect. Sure, it's still touch, but better than most Linux touchscreen input methods I've used. I even use it on my Steam Deck today. That is a great idea.

KDE Connect is super solid for that. Thank you, Jeff. I use it all the time on my media, sort of watching, you know, when you're on the couch and you don't feel like getting up to do anything. It's actually been great. So good tip. I hadn't even considered pairing my phone to the Steam Deck with KDE Connect. I'm doing that tonight. As for computers, Jeff continues, running in hot environments, I have my Rock Pro 64 in my garage as my drumming reaper machine. It's typically off, and I try not to drum when it's

over 100 degrees Fahrenheit in there, but I most definitely have recorded in 110 degrees. The Rock Pro 64 doesn't even have a fan, just a passive metal case, and it gets too hot to touch. But the CPU never gets above 65C with a dozen Reaper tracks playing and a few recordings going. I'm just getting a mental picture of Jeff just wailing on his drums in 110 degree California heat. That is probably a stinky mess, to be honest. Well, and Jeff, you gave two numbers here in two different standards for reporting temperatures. And so I am unfortunately going to tell you that 65 C is 149 Fahrenheit. So that's pretty hot.

Yeah. Yeah. Doesn't your skin melt off your bones at that temperature? Not Jeff. You know, he's pretty used to it. He's, he's, uh, he was born in the sun. All right. We had a 19 boosters this week. Thank you everybody. There is a last minute boost. Oh yeah. Really? Yeah. Okay. Are you sneaking it in right now? I sure am. Nev boosts in. It's a zip code boost. This is a hot boost.

That's also the name of a movie. It's 45,365 sats. Coming in hot with the boost. Just a reminder that Ohio Linux Fest is less than two weeks away, and I plan to be there. Join us in the Columbus Club in Matrix. Yes. Thank you, Nev, for the reminder.

Really appreciate that. We always are up for you mentioning it. We just sometimes don't have it on our radar. And 45365 is a postal code in Ohio. Sidney Kirkwood or Hardin. Hello, Ohio! Did you say there was a name of a movie in there? I don't know that movie.

Yeah, it explores the congruities of daily life in an American town. From the patrol car to the courtroom, the playground to the nursing home. Noted. A little homework for you, Brent. Thank you. you thank you everybody who boosted and we had 20 total boosters now thank you nev 24 boosts and that was a grand total of 619 701 sats it really whips the llama's ass thank you everyone uh we love the interaction here it gets us talking about stuff we never planned to talk about. It's the most spontaneous moment in the show for us,

but also it's a great way to just send back a little value that you got while listening. And you can do it by switching podcast apps if you're ready for a brave new world. You can go to podcastapps.com. Fountain, Castomatic, and Podverse are the most popular in our community. Or if you want to keep your dang podcast app, just get Albi. You could do the experiment without switching apps. Get Albi.com, top it off either directly or using something like the Cash app. That's what I use because the Lightning Network is just an open network.

You just send sats into Albi. Then you go to the podcastindex.org, find Unplugged, and you can boost in right from there. Once you get it going, it's actually really straightforward. It's that initial journey that's just a little treacherous sometimes. But we thank you for trying to make it. Now, we could have made the show out of this pick. We really, really could have. But it's going to be a homework project for the listener. Our pick this week is called Llama GPT, L-L-A-M-A, GPT. And it is a self-hosted offline chat GPT-like chatbot powered by Facebook's open Lama 2 engine.

No data leaves your device. It all runs, and they've optimized it to run on consumer-grade hardware. Now, I'm not saying you're going to be blown away by the performance, but they've actually tried to make this thing even usable, depending on the model that you pick. They've tried to make it even usable on the raspberry pi which is incredible yeah sure it's a 0.9 tokens per second compared to 54 tokens per second on a m1 max but hey that's almost one what they're really doing here so this is umbral you guys we might remember them we've mentioned them before

they make a bitcoin node with a with like an app marketplace and they dockerize up these apps and what they've done is they've kind of released this for general consumption and they've just made available with a really simple docker compose or whatever you want to however you want to do it and it'll all spin up it'll download the models for you and they've even just slipped in that new code llama that facebook released uh yeah you know previously code generation specifically had kind of been one of uh llamas not so great, but now there's a more specific support in this new code Lama model. That should be fun to play with. I also see that thanks to some of the underlying tech like Lama.cpp, this thing also offers an open AI compatible API.

So maybe you've already got some tooling that sort of is hooked to integrate into the open AI world. Well, you can use this to redirect that to a Lama model. And that.cpp is important because it means it's designed to run on your CPUs. So you don't have to have yourself an NVIDIA GPU farm to play around with this stuff. You really just need Linux with Docker or even macOS with Docker. And you can run this. And they've stacked a bunch of really good open source projects together like my favorite chat gpt front end is what's being used to communicate with all this but i think probably the more important thing about this pick

is that it is bringing the open source alternatives to open ai to all of us from a raspberry pi up to a laptop i'm running on my desktop workstation long term i'm really kind of the most interested in these open source large language models do you agree yeah i mean they're the you know they're the ones you can and we should be clear there's some you know uh concerns qualifications around just how open llama is but you can just see by the explosion of, you know, stuff and open source projects and packaging and wrappers and tools on top of it already, kind of like what happened with stable diffusion,

you know, when these things are more open, that just spawns the ecosystem to build all these gadgets and you can, you know, you don't have to worry about the expense so much. You need hardware to run it, but you're not constrained or limited by whatever the API gatekeepers want you to be able to do. The nice thing about this Lama GPT pick as well is they seem to be really following the developments of

the open source tooling and from the front ends to the back end models and they're making it available really easily to just stay on that and keep checking it out. And I think I want to suggest if you have time, you should try one of these because i think it helps give you perspective of how these really are just tools they're not they're not super dangerous and they're also not extremely capable either but they are useful and they let you get under the hood here and tweak a few things and you can get something that's more concise or something that's more quote-unquote creative and you can really kind of get a better understanding of how the machine works and i think

anytime you have an opportunity to wrap your head around tooling that there's a lot of a lot of hype about a lot of concerns around it's a it's great to just educate yourself in a way that is self-hosted local totally private and taking advantage of the most modern open source stuff we get to play with. Yeah, right, especially with these kinds of prediction and inference things. There's a lot of magic and black box stuff going on just in what it is, and then when it's hidden behind a proprietary API,

it just takes that to the next level. And if these things are going to be integrated into so many parts of our lives as we keep being, as is suggested and we've been seeing to some extent, yeah, I think it's kind of pivotal that we understand at least at the high level of how they work and what they can do. I think, too, this is the first step into these maybe one day community hosted large language models.

You know, I was talking about this on Office Hours, but what if you went to notes.jupyterbroadcasting.com instead of just a text search, there was an embedded large language model search. And you could say, what episodes did the guys talk about Graphing OS? And it would not only find you the time codes, but maybe even give you a little summary and context of what we have said about Graphing OS in one concise spot. Then, next layer reads you the answer in the AI Chris voice.

Oh, man. Let's be honest. Everybody would the AI Chris voice. Oh, man. Let's be honest. Everybody would pick the Brent voice. That's what I would choose. I'd have him read me a nighttime story. Well, you can ask that. I would do that for you. But you're right.

I mean, like, kind of having access to these things, we might actually want to do that this way. But if it meant, like, we had to budget out calls to open AI's api or something like that's just not gonna happen we're just it's not gonna be on the priority and it's not as accessible to the community to try to experiment with the tooling because they also have to have an open ai subscription and pay them 30 bucks a month or whatever it is and just it's a barrier and then at the end of the day they can pull the plug anyways and so i think we've learned our

lesson there right well let us know what you think about that give it a go and uh report back into the show it'd be a great excuse to boost in or go over to linux unplugged.com contact because of course we will have a link to llama gpt and everything else we talked about in the show notes today why not ask it to write us a nice boost message they can help you out and then come live why not have the ultimate experience we do it sundays at noon pacific 3 p.m. Eastern. See you next week. Same bad time, same bad station. Although most of you just prefer to download it. I get you. That's how I get my podcast. RSS feeds are at linuxunplugged.com slash RSS, or just find all the different direct podcast app links over there. And of course, the feedback pages there, the links, everything we talked about, it's all stacked over there,

nice and easy. And then if you want the next level, go to jupiterbroadcasting.com there's a whole suite of fine shows over there there'll be a fresh coat of radio coming out just a little bit after this episode get a great take over there as well but that's it for us, thanks so much for tuning in this week's episode if you do play along and try out LamaGPT or Asahi Linux please write in, let us know how it goes. We'd be really interested. Thanks so much for joining us and see you back here next Sunday. Thank you.