Untitled Linux Show 276 Transcript

Please be advised that this transcript is AI-generated and may not be word-for-word. Time codes refer to the approximate times in the ad-free version of the show.

Jonathan Bennett [00:00:00]:
This week we're talking about what's new in OBS, in Calibre, in Sonic, and a quick detour going from ext4 to ZFS. Then there's PipeWire, there's more file system news, and then the big stuff. Linux kernel 7.3 is out. Umarchi has an interesting update. There's a de-Googled Android phone that might be worth taking a look at. And it was a very bad, rotten, no good week for security. All that and more, you don't wanna miss it, so stay tuned.

TWiT.tv [00:00:34]:
Podcasts you love. From people you trust. This is TWiT.

Jonathan Bennett [00:00:43]:
This is The Untitled Linux Show, episode 276, recorded Saturday, October the 24th. Didn't have enough dice. Hey folks, it is Saturday and it is time to geek out about Linux, open source software, probably some hardware thrown in there for good measure, and all kinds of other stuff. It is the Untitled Linux Show. I am your host, Jonathan Bennett, and I've got the cadre, the crew, the trio, the tres amigos. I guess I'm the fourth amigo. So los cuatro amigos? Just doesn't roll off the tongue quite the same way.

Jeff Massie [00:01:16]:
The four amigos.

Jonathan Bennett [00:01:17]:
The four amigos. That one works better in English, doesn't it? Anyway.

Jeff Massie [00:01:21]:
Four Musketeers.

Jonathan Bennett [00:01:24]:
That could work. That could work, I suppose.

Jeff Massie [00:01:26]:
UL Quattro sounds like a car.

Jonathan Bennett [00:01:29]:
It does sound like a car, doesn't it?

Rob Campbell [00:01:30]:
Quattro.

Jeff Massie [00:01:31]:
Did you see my new ULS Quattro? Oh, it does great in the snow and it's—

Jonathan Bennett [00:01:35]:
Anyway, there's some stuff that has happened this week. There's a lot of stuff that's happened this week, actually. It's been kind of a— In fact, let me just say my last topic is labeled the no good, rotten, the very bad, rotten, no good week for security. We'll get to that here after a bit. And yeah, that one's interesting.

Jeff Massie [00:01:57]:
Is that because it was boring?

Jonathan Bennett [00:01:59]:
No, no, it was not boring. And it's

Rob Campbell [00:02:01]:
been all over the place, so everyone should know about it.

Jeff Massie [00:02:05]:
My first story is not even on the news feeds.

Jonathan Bennett [00:02:09]:
Well, the first

Jeff Massie [00:02:10]:
story—

Rob Campbell [00:02:11]:
Neither is Jonathan's.

Jonathan Bennett [00:02:12]:
Neither. Well, so one of mine is. This is kind of a 2-parter. So yeah, I'm going to talk about OBS, the Open Broadcast System. And there's 2 things, 2 things to note here. One, they just released OBS 33. I think we talked about this a little bit in the beta. It's got some cool stuff in it.

Jonathan Bennett [00:02:27]:
They're continuing to work on plugins and they did a big update for CEF, C-E-F. That's the Chrome Embedded Chromium, excuse me, Chromium Embedded Framework. And that's the virtual browser that lets you do things like load up. Well, if you play with Video Ninja, that uses CEF to do the actual capture from the browser. If you want to do Any of those things where you're capturing from the browser, CEF is the framework, the part of OBS that lets you do that. And then of course, they've got updates with NVIDIA video encoding and some ARM64 stuff. So there's neat things in OBS 33. But there is a thing that I have wanted in OBS for literally years now.

Jonathan Bennett [00:03:16]:
I have ever since, um, ever since PipeWire came out and became a thing and solved Linux audio, in my opinion, just fixed it. Um, and then the PipeWire guy said, we're going to do it for video too. And I immediately went, yes, I can finally get rid of that old v4l2loopback thing. And I can have OBS output directly into a browser or anything else to be able to do a virtual camera. And I knew it was going to happen. And then a couple of years ago, there was a pull request opened in OBS that made it work. But yes, finally. And the OBS guys said, no, we're not going to pull that.

Jonathan Bennett [00:03:58]:
Like, and the reason, of course, was some security thing using OBS inside of a Flatpak. The desktop portals didn't have support for it. And my opinion is that's a really dorky reason because this is a really narrow problem and it would work for everybody else. And they should have just— Hold the dang patch and let us have our fun. The world has changed. And this is actually really interesting, kind of a bigger point. But the world has changed in the last 2 years, and we now have access to artificial intelligence, or large language models, if you prefer, or superintelligence, if you prefer. Not a huge fan of that term.

Jonathan Bennett [00:04:40]:
Anyway, I have a Claude subscription, and it occurred to me the other day that Claude can probably compile this for me with this patch in and maybe some other things that I've wanted to do for a long time. So I said, hey Claude, fork this for me, pull this patch in and make it work. And about 30 minutes later, we have an OBS that is working with Pipewire to shuttle video into a I mean, I don't even know that you would call it a virtual, like it says it's a virtual camera, but it's kind of not even that. It's just a PipeWire pipe. So I am, you know, I'm inside of, here, I'm inside of OBS. I can even do this. Let's see, we're gonna change the screen capture from that to— OBS. That.

Jonathan Bennett [00:05:34]:
There we go. Look, it's OBS. And, uh, you can't quite see it because of where my— I am at. I can fix that too. I added this as well. Um, if I can find it yet. Transform. We're gonna say, uh, bottom left corner.

Jonathan Bennett [00:05:52]:
There we go. And then we've got the virtual camera here running. And, uh, it is— Yeah. Output type default. It's just doing its thing. You know, it works. It's boring. It's the best kind of programming.

Jonathan Bennett [00:06:09]:
It just works and it's boring. So I have forked OBS and I've pulled the patch in. I've made it work. I've added a couple of other things. As I need things in OBS, I'm just gonna ask Claude to do it. Now, so one, that's cool. And if you want this, feel free to go and, you know, download OBS from my little fork, you know, my one-feature fork. But it brings to mind like this next thought that the way we do these open source projects have kind of fundamentally changed because we're no longer— and I have a couple of thoughts here.

Jonathan Bennett [00:06:49]:
It's fundamentally changed because rather than— well, it's lowered the barrier to entry so much. And so rather than I've got to download OBS the way that the guys that run OBS want it to be used, I have to use their vision. Like, there's nothing necessarily wrong with that, but I have a different use case that I wanted to support. And it is now extremely easy to change it and support that use case through an LLM because you can literally ask Claude to do it and 30 minutes later you have the working thing. And that kind of fundamentally changes the way that this open source thing works, because now everybody can run, and I've seen this in other projects, I've seen this all over the place, everybody can run their own customized thing that's based loosely on the upstream, which really changes. Again, I've been thinking about this a lot this week. I'm like, it really changes the sort of the agreement. And we're now, you know, as an end user, I'm no longer, tied to the vision, the choices that the upstream made.

Jonathan Bennett [00:07:57]:
And then the other really interesting thing with that, so you might ask, oh, have you pushed this? You fixed the patch, right? Have you pushed it upstream? No, because the OBS project has a strict no AI code policy. And so they don't want it. And in fact, I had to tell Claude while I was working on this, I'm like, Claude, go look at this file, see their policy, don't open any pull requests, don't tag anything, don't make comments. They don't want you there. They don't want your kind in the OBS project. But I made it work. And I'm pretty happy with that. I also find it interesting though, because if you think about like the philosophy of open source and sort of the original Free Software Foundation, the freedoms, One of the freedoms that they believe very strongly in is the ability for people to modify the code that ran on their own computer.

Jonathan Bennett [00:08:56]:
And, you know, we're kind of moving into an era where we have that freedom in a bigger way, I think, than they ever imagined it. It's at least accessible to more people, which is cool. It is challenging in some ways. It causes some problems in some ways, but it's also really exciting and it's very cool. So that's my OBS story. And I am happy. And I got to say, when I made it work, like 30, 45 minutes of working with Claude finally made it work. And at the end of it, I'm like, I got a little emotional because I've wanted this for like years.

Jonathan Bennett [00:09:32]:
And then finally I realized I could do it. And 30 minutes later, it was on my desk. It's like, wow.

Ken McDonald [00:09:38]:
I've actually been running a personalized version of G Potter.

Jonathan Bennett [00:09:43]:
Do you have— did you have the, uh, the clanker do that for you?

Ken McDonald [00:09:48]:
Gemini helped me with configuring it so that I could get the, uh, you, uh, down the YouTube DLP to work with it without having issues and having to worry about trying to— yeah, any updated versions back in.

Rob Campbell [00:10:10]:
Cool. So I was going to avoid talking about AI on this show because I received so many complaints, but Jonathan just went all in right at the beginning. And

Ken McDonald [00:10:20]:
I didn't say anything about artful impatience so far.

Rob Campbell [00:10:27]:
I meant superintelligence.

Jeff Massie [00:10:30]:
I mean, I used it. I was talking before the show started. I used it to help set up a mini DNLA server to stream to a little box so I can play music, bit-perfect music to my stereo. And it went, oh, here, let me help you set this up. Oh, you got a firewall. Here's the command you need to run as sudo, because it doesn't have my sudo password, so that you can open the ports just for your local LAN so that it's achievable. And it— You know, it was like, oh, this took just a couple minutes to— when I'd been fighting with it myself for a little while.

Rob Campbell [00:11:08]:
And, but it's, it's like I've been saying, you know, if, if you aren't going to listen to your customers, they're just going to do it themselves now. And, you know, I know there's been plenty of times, work-related or otherwise, where it's like, I have this thing, it's A very specific one was like, I have this thing, it's like 20 years old. I really need an updated version. Nobody makes anything decent to replace this. And I had AI just make me something. I was like, you know what? Fine. No one's going to support my needs and what I want and need. I'll just make it myself.

Rob Campbell [00:11:50]:
Which has always been the open source programming way. make it yourself, but now it's lower barrier to entry.

Jonathan Bennett [00:11:55]:
Yeah.

Ken McDonald [00:11:56]:
And with my birthday present that I showed off a couple of weeks ago, I've already seen modifications to the firmware for it that I'm tempted to upload into—

Rob Campbell [00:12:09]:
Your InkBook thing.

Jonathan Bennett [00:12:11]:
Yeah.

Ken McDonald [00:12:11]:
Yep.

Jonathan Bennett [00:12:12]:
Oh, I'm sure. I'm sure people are going nuts with that because it's probably got simple enough firmware that, yeah, it's right in the target zone for using an AI to work on it.

Ken McDonald [00:12:21]:
You know, this kind

Jeff Massie [00:12:21]:
of almost reminds me of an analogy of back When I first got on the internet, this was about '91 or so, if you didn't compile your own stack, you didn't really deserve to be on the internet. AOL was horribly looked down upon because you didn't have to work for it. And I look at it almost like the open source software now. It's like, oh, you didn't program it yourself. You had the easy way. Therefore, you know, It's kind of, I kind of see a lot of similarities there.

Rob Campbell [00:12:56]:
So, so Ken, how does the InkBook work with Calibre?

Ken McDonald [00:13:00]:
Good. I've got it set up. I just did a customized device configuration and it's working great. But I'd like to tell you about some of the new features that Calibre itself has. Especially since the last time I even covered Calibre was back in August of last, last August, where I'd covered over the updates up to version 9.12. As of yesterday, COVID Goyle released Calibre version 9.17, adding over 20 new features since 9.12. Some of the notable features include the ability to generate books using book covers using AI with various styles such as romance, sci-fi, and pop pulp fiction. Now in the edit book section, when you're hovering the mouse over an image file, a floating preview of the image will now appear next to the file browser.

Ken McDonald [00:14:02]:
And a new command line interface flag that allows users to add or delete a book from a specified path in the system's directory while running Calibre.exe is available. There's also a create-your-own-adventure writing game featuring interactive storytelling with AI managing the created world. And having played with that, probably won't. Now, support for downloading news from websites that block automated access using the CamuFlox browser has also been added. Now, one feature I have played with is the read aloud feature. It has now been enhanced with the addition of highly high-quality Kokoro voices, which sound much more natural than the previously, previously used Piper voices, particularly for English. Actually was listening to the Linux Pocketbook that way the other day. Or Pocket Linux Book.

Ken McDonald [00:15:06]:
Support for reading images in the JPEG/XL format during format conversion has also been added. And there have been improvements to the Edit Books checkbox feature, which now includes a check for dangling accessible rich internet applications. I'm just going to say ARIA for any future references to that. But for those references, and it will also update invalid IDs when auto-fixing. And the last one I'm going to touch on is compression speed has been enhanced, making conversions to MOBI, AZW3, and PDB output formats faster. I miss my Palm. Now, according to Bobby, the conversion engine in Calibre 9.17 now supports CSS Level 4 selectors, which enhances ability to process more modern styling rules found in eBooks. Markus also reported that the content server and eBook viewer received, received several improvements, including notifications to the Calibre GUI about new formats created by conversions in the content server, as well as various other fixes.

Ken McDonald [00:16:27]:
Now, since I only touched on a few of the improvements, I do recommend reading Bobby and Marcus's article for more details. And you can also visit Calibre's What's New for all the improvements in versions 9.13 through 9.16. I keep wanting to trip over that dot.

Rob Campbell [00:16:49]:
Hold up your hands, Ken. Your palms look fine. I don't know why you miss them.

Ken McDonald [00:16:57]:
Uh, uh,

Jonathan Bennett [00:16:58]:
bravo.

Ken McDonald [00:16:58]:
I'm trying to think. I started with the PropBomb M150.

Rob Campbell [00:17:03]:
I think that, uh,

Ken McDonald [00:17:06]:
uh,

Rob Campbell [00:17:06]:
that's the create your own adventure, the your own adventure thing. That sounds interesting, at least.

Ken McDonald [00:17:12]:
So you may, uh, install Calibre just for that?

Rob Campbell [00:17:16]:
Yeah, I don't know if I'd use it a lot, but at least to check it out, that, that, uh, it seems something interesting just to see how it works.

Jonathan Bennett [00:17:26]:
Yeah.

Ken McDonald [00:17:26]:
Now that I've got this new ebook reader, I'm using it at least once or twice a week as I finish books and transfer them off and on.

Rob Campbell [00:17:37]:
Only once or twice a week? You read way faster than that. Come on.

Ken McDonald [00:17:42]:
It holds more than one book.

Rob Campbell [00:17:44]:
Oh, gotcha. You do more than one book at a time. That's a whole pack.

Ken McDonald [00:17:48]:
In fact, thanks to Humble Bundle, I got a collection of, uh, Corey Doctorow, uh, books that I'm going through right now.

Jonathan Bennett [00:17:59]:
Yeah, cool. I enjoy some of his work. All right, let's, uh, let's talk, uh, let's talk plasma, Rob. You're, you're, um, you're taking a step away from your, your normal character here.

Rob Campbell [00:18:13]:
I am.

Jonathan Bennett [00:18:14]:
I'm doing that. Breaking character. What's, uh, What you got for us?

Rob Campbell [00:18:19]:
So, you know, I know some of you love KDE Plasma, but really hate Wayland. And now that KDE has decided to drop X11 support with Plasma 6.8, well, you're probably not very happy about it at all. Personally, I think it's time to get over it and move to Wayland. But hey, That's not my decision to make for you. And this is Linux where you don't, if you don't like the direction a project is going, like OBS for Jonathan, there's a pretty good chance somebody will fork it. And that's exactly what happened here. Let me introduce to you Sonic DE, the Sonic Desktop Environment. Sonic DE is a community-driven fork of KDE Plasma 6 that aims to preserve And improve X11 support rather than abandon it.

Rob Campbell [00:19:16]:
And this isn't just somebody maintaining an old version of Plasma. The developers are actively working to create a complex desktop environment, removing Wayland-specific dependencies and continuing development of the X11 components that KDE is leaving behind. The project actually started back in July of 2025 as a collection of improvements to KWin's X11 Windows Manager. It eventually grew into a project called KDE Lite before being renamed to SonicDE in December. I kind of like the KDE Lite, but yeah, whatever. Maybe there's some conflicts there with the naming. And it's grown quite a bit since then. According to the project's website, they're now maintaining 139 components, including their own Windows window manager and compositor called SonicWin, desktop workspace components, a login manager, and even their own Silver desktop theme.

Rob Campbell [00:20:17]:
They're also working on a Vulkan backend for their compositor, and they want to support both the traditional X.Org server and the new X-Libre project. Anyway, another interesting aspect is that SonicDE wants to remain independent of the init system. So unlike modern desktop environments that are increasingly tied to systemd, SonicD wants to remain compatible with alternatives. Basically, they're trying to just stop everything from moving forward, I think, here. But anyway, they're even looking at supporting BSD operating systems. Now, why would somebody want this? No, serious. But there are still users who have legitimate reasons for preferring X11. You know, maybe it's that compatibility with older applications, certain remote desktop tools, global keyboard shortcuts, workflows that don't behave quite the same way under Wayland.

Rob Campbell [00:21:15]:
Personally, I've been using Wayland for years and I think it's, I mean, it's not even the future. It's the present of the Linux desktop. But I also appreciate that not everyone's hardware workflow is the same and there's niche cases. So, Here's another niche case desktop for you. And that's what makes projects like SonicDE interesting. It's another example of the flexibility of open source software. KDE is free to move forward with Wayland and SonicDE developers are free to take the existing code in whatever different direction they want. That's open source.

Rob Campbell [00:21:50]:
Now, whether SonicDE can maintain a complete desktop environment over the long term is another question entirely. That's a pretty substantial undertaking, especially as KDE continues to develop Plasma. I imagine it's going to be like Cinnamon and GNOME where, you know, one goes one way, one keeps going another way, and soon they're just way different. But for those of you who love KDE experience and absolutely refuse to give up X11 from this hand, SonicD might be worth checking out. You can find the project, its source code, and installation instructions at sonicde.org. If I'm not saying that clear, that's Sonic as in S-O-N-I-C-D-E.org. And I'm curious, are any of you still holding out on X11? I know you are. And if so, why? What's keeping you from moving away, Lynn?

Jonathan Bennett [00:22:52]:
If you're hanging on to X11, what's wrong with you? I mean, I feel like that list of legitimate reasons is shrinking, but yeah, I'll give them the benefit of the doubt.

Jeff Massie [00:23:05]:
If it's good enough

Ken McDonald [00:23:06]:
for

Jeff Massie [00:23:07]:
S390, it's good enough for me.

Rob Campbell [00:23:10]:
There's always XFCE.

Ken McDonald [00:23:12]:
And the list of security-related reasons to move away is growing.

Jonathan Bennett [00:23:19]:
Yeah. Yeah. Although—

Rob Campbell [00:23:20]:
I say that all the time, but I'm going to be nice today.

Jonathan Bennett [00:23:22]:
Although, not to make this the Untitled AI Show, but, and I think I made this point a week or two ago, but it's not actually that difficult to think that now a single developer can keep up with the security load for something like X11. Just have the robot fix the things.

Rob Campbell [00:23:42]:
Look at how far we've come.

Ken McDonald [00:23:44]:
As long as they stay interested, right?

Jonathan Bennett [00:23:46]:
Yeah.

Rob Campbell [00:23:47]:
Look at how far we've come on this show. Do you remember when Jonathan was the all-out anti-AI guy and the rest of us mostly weren't? And now we're trying to keep quiet and not even talk about it, and he just keeps bringing it up. Oh, I don't know that.

Jonathan Bennett [00:24:01]:
I don't know that I was the anti-AI guy. I was skeptical of it, but it got good enough and I started using it and I just couldn't go back.

Rob Campbell [00:24:09]:
Just keep bringing

Jeff Massie [00:24:10]:
it up. I don't know if you could really Do that though with AI and X11 because it's fundamentally kind of broken to be secure? I mean, it's going to have to— it's going to be some heavy-duty code lifting. And then at what point do you go, well, it's not really X11 anymore because you had to rewrite so much of it?

Jonathan Bennett [00:24:30]:
I, yeah, I don't, I don't mean fixing the, the fundamental things. So that would be, that would be essentially doing the same thing that Wayland did, that, that kind of a rewrite. No, I just mean keeping up with the found CVEs. I think, I think a one-man band with, you know, a good subscribe— you know, a good LLM subscription could keep up with the found CVEs. Because at one point it was, it was unclear that X-Ray was even going to be able to do that because things were continuing to be found. And the engineers that were actually working on X11 were all paid by Red Hat. And Red Hat has, if they have not already, they are about to stop doing that. Because the oldest version of Red Hat Enterprise Linux is now dropped out of support, and it's the one that actually used X11.

Jeff Massie [00:25:18]:
I guess I'm saying with it being fundamentally designed wrong for today's hardware and environment that the seemingly— you're only seeing the tip of the iceberg because there's so many other fundamental things wrong that It takes more than just a few, some patching. It's rewriting.

Jonathan Bennett [00:25:39]:
Yeah. And that kind of depends upon where you stand on, you know, are those security weaknesses actually vulnerabilities or are they understood to be—

Ken McDonald [00:25:51]:
Undocumented features?

Jonathan Bennett [00:25:52]:
No, not even undocumented features, but like the question becomes, where is the security line that you're going to draw? And with X11, it was, You know, we can run this as root and the person running the computer is trusted. And so therefore we kind of trust the rest of the programs that are running. And with Wayland, they took a different approach. And it's like, we don't trust the other programs that are running. We also want to protect against those. Don't trust anybody. Yeah. So it's a different security posture.

Jonathan Bennett [00:26:22]:
And so long as you're okay with the security posture of X11, then it's maybe not as fundamentally broken. But if you follow—

Rob Campbell [00:26:31]:
In reality, there are plenty of use cases where the user of it is trusted, the application is trusted. You know, you're air-gapped and just doing some little thing in some random corner of whatever random part of the world.

Jonathan Bennett [00:26:46]:
Yeah. So, all right. We are going to, we're going to take a quick break and then we're going to talk about file systems, at least one file system. We got a lot of stuff coming up, but don't go anywhere. We'll be right back.

Jeff Massie [00:26:59]:
So this is, this is my own. There's no links in the show notes for this one. This is my own personal adventure. And for you, you know, hardcore sysadmins, this is going to be nothing. But for the newer people, and this was my first time through this, is here, here's my little adventure. So now I have a home server for all my documents and media, and I run Plex on it, and I played with Jellyfin and stuff like that for serving my media. Uh, you know, I've had it for years. It runs on 3 18-terabyte drives in a Linux software RAID 1 mirror using mdadm with plain old ext4 as the file system.

Jeff Massie [00:27:42]:
You know, 3 copies of everything, belt suspenders and a backup pair of suspenders. And, you know, it's worked great. Well, this week I moved All the, you know, I had 10 terabytes of media. I moved it all over to ZFS, and basically I just kind of want to walk you through how I did it and, you know, how I kind of kept things safe by without ever having fewer than 2 copies of my data. Now, the first thing is probably why bother? Why would I even go through this if my current system's been working for years? Well, MDADM is rock solid, and I've done stories about it in the past using MDADM and, you know, how it's actually more stable than a lot of RAID hardware, you know, until you get to the enterprise-level stuff. But for most consu— for the consumer level, even semi-pro, MDADM has been around for, you know, a very long time, and it is very trusted. But there's an issue. MDADM mirrors blocks without knowing whether those blocks are correct.

Jeff Massie [00:28:52]:
If one disk quietly hands back bad data, you know, which a lot of times we call bit rot, MDADM can't tell which copy is the right one. And ext4 can't tell you either. Now, with 3 disks, it'll say, oh, We have one mismatch, the 2 other agree, okay, that must— the one not matching must be the wrong one. But if you have just a regular mirror where you just have 2 disks, it doesn't know at that point. Well, ZFS checksums every block it stores. So when it reads a block that doesn't match its checksum, it pulls the good copy from another disk in the mirror and repairs the bad one automatically. And on top of that, you can, if you want, get snapshots, compression, per-folder settings. You know, it's got a lot of features.

Jeff Massie [00:29:46]:
And for those that don't know, ZFS has been around for decades. It's a very proven file system. Now, I will say, you know, for people that wondered, I did look at BTRFS. You know, BTRFS mirrors are fine, but You know, someday I might want to go, you know, having a parity RAID, meaning RAID 5 or 6. And the BTRFS documentation still says don't use it in production. The problem is called a write hole. So you lose power in the middle of a write and the parity can disagree with the data. And there's— this has been looked at.

Jeff Massie [00:30:26]:
There's a patch series that was posted in June, which supposedly fixes it with something called the— RAID 5. RAID stripe tree, but still to this day the author says, you know, it's not meant to be used yet. So it— so ZFS, you know, is what is where I went. Uh, with 3 drives I could have used RAID-Z1 and doubled my speed, but at this point I just stuck with a 3-way mirror. You know, some You know, I, I felt good with what I had and I said, I'm just keeping it. Now there's one catch to this, and it's there's no in-place conversion from MDADM to ZFS. You have to copy everything. And so I'm— I had to, you know, I didn't have even another 18 terabyte drive or greater laying around, so Here's, here's what I had to do.

Jeff Massie [00:31:25]:
Because I had a 3-way mirror, I had a disk to spare, so you borrow one. So step 1 was I tell MDADM that one disk has failed, remove it from the array, and wipe it. The, the MDADM array keeps running on the other 2 disks. You know, they're still mirrored, so I still have 2 good copies. And then the second step is I build a ZFS pool on the disk that I failed, that I freed up. I happen to name mine tank, which is the traditional ZFS pool name. You know, it's kind of one of the— a lot of people use it. It's kind of like naming your dog dog, you know.

Jeff Massie [00:32:05]:
Now, I won't get into the tank meaning, such as storage pool of water or fish tank, or there's some historical thoughts of it's tied to a character in The Matrix and But I'll let you figure that one out, Mr. Anderson. Now, there are a few settings when I first made that pool that matter here. The first one is I set a-shift equals 12. Now, that tells the ZFS that the drive uses 4K sectors. Now, these drives claim 512-byte sectors for compatibility, but they actually use 4K internally. And if you use 512, it works, but there's a lot more churn on the disks. So it's ideal to use what the internal drive— the, the sector size of the actual drive.

Jeff Massie [00:33:02]:
Uh, now the thing here is you can't change a shift later, so you want to spend a little time, get it right up front. Now, I also turned on LZ4 compression, which is, you know, it's pretty cheap to use, you know, CPU-wise, and a lot of times it can be faster to have it on. And it, uh, if it runs into stuff it can't compress, it skips it. So it's not like it's trying to churn on a lot of like media files that it just is not going to have any luck with. I also set ATIME to off— that's access time. So when I have Plex just playing a movie, it's not going to trigger a write to say, oh, let me update the time that file was last accessed. I didn't need that. So then I, I split my directories into datasets.

Jeff Massie [00:34:04]:
One was for media And then the other one, you know, it was, it was a 1 megabyte record size for the, for the compression. And then I had a smaller one for documents, and that got a lot more out of the, the Zstandard compression. Now, as a kind of a confession here on myself, I assume ZFS was already installed because the system reported ZFS. version 2.4.1, and this is on my Kubuntu 26.04. Well, that's the kernel module which Ubuntu ships in the current— in the kernel. The zpool and ZFS commands live in a separate package called zfs-utils-linux, so I had to apt install that the first time. You know, learn from me. Just be aware there's, uh, there's some steps you have to do.

Jeff Massie [00:35:00]:
Now, step 3 is the big copy. So I did an rsync. I did a -AHAx, and basically that's archive mode plus hard links, ACLs, and external attributes. So everything comes across exactly as it was. Now, it takes a while. I've got spinning rust drives. It took a little, roughly 12 hours to do it. And kind of another little hint, and we've talked about this on the show before, is use tmux.

Jeff Massie [00:35:36]:
So I was on my main computer and I was just logged in remotely into my server, and tmux keeps your session alive. So even if, say, you're on a laptop and it falls asleep, or I reboot my main computer, my gaming computer, tmux keeps that going. So you want to make sure that you don't have something that's going to kill your rsync. Okay, now step 4 is where I did a switchover. Anything in this case that— well, anything that might be accessing the MDADM RAID drives, I turn off and I switch over to the ZFS drive. So it's only having— it's accessing just that one copy, but I still have the 2 MDADMs in reserve. Now, there's also a recommendation if you look at a lot of the online documents, run rsync again later to make sure that everything still matches. I didn't have to because mine being a media, mostly media, it's not getting a lot of writes and there's not a lot of churn on the drives and I knew Because I'm the only one accessing it, that it was fine.

Jeff Massie [00:36:51]:
So basically, I did it again. You you take another drive and you fail it. Now you add the ZFS that that next failed drive and you add it to the ZFS pool and. That's where it starts what it calls a resilver, which is basically rebuild. Now, unlike MDADM, which rebuilds the entire disk, ZFS only copies blocks that actually hold data. So it is a little faster than an MDADM rebuild, but, you know, it— for in my case, I had 10 terabytes and it took about 14 hours and 9 minutes. Um, you know, now you're— be aware that your drives will get warm. I had my drives peaked at about 50 degrees Celsius, uh, under— underneath the 60-degree thermal limit, and that was for the whole rebuild.

Jeff Massie [00:37:55]:
But just be aware that— make sure there's some airflow to your drives because they're going to be running hard for a significant final time. And the final step After that, that was all good. Everything was mirrored, everything, no errors. Then I had to stop MDADM for good, wipe the last disk, remove the array from the MDADM config, rebuild IntraRAMFS, and attach the disk as the third in the mirror. And it, it added that third to the pool, rebuilt it again. Every— no, no, uh, errors. Everything went great. And now I'm running on ZFS, and now I have the block level correction that I need, or that I wanted, because I, you know, a little paranoid about stuff.

Jeff Massie [00:38:52]:
Um, so that was just kind of what I did. There's a lot of online guides, you know, AIs can help you as well. Make sure you don't execute any commands you don't know what they're doing. I had some mixture of AI and online guides, and I would make sure I knew what every command was doing before I executed it. And it will take, you know, depending on the size of your data, whatnot, it could take a few days to do. But as long as you're careful and very thorough about it, it can, it can get done, and then you'll have a much more stable RAID. It is not a full backup, but it is a RAID. So have fun RAIDing the computer.

Jonathan Bennett [00:39:39]:
I learned that the hard way one time, that RAIDs are not backups, very early on in my career. Yeah.

Ken McDonald [00:39:48]:
They're just a way of ensuring the integrity of your data.

Rob Campbell [00:39:53]:
They're hardware redundancy.

Ken McDonald [00:39:54]:
It's used as the primary storage.

Jonathan Bennett [00:39:56]:
Yeah, hardware redundancy is a good quick way to put it.

Jeff Massie [00:39:59]:
I would say, yeah, they're part of a good backup solution.

Jonathan Bennett [00:40:03]:
Yeah. Now, what distro are you using and was it difficult to get ZFS support in it?

Jeff Massie [00:40:11]:
It was Kubuntu 26.04. So it was—

Jonathan Bennett [00:40:14]:
Does it ship with ZFS?

Jeff Massie [00:40:17]:
Yeah, well, that's where I said, learn from me. It ships with the kernel modules already built in. But it doesn't have all the tools. So you have to install the tools, but it's in the repository. So it's just an apt install and it's in there.

Jonathan Bennett [00:40:30]:
Isn't it ZFS that the license of it is weird because it came out of Sun? Yeah. Yeah. That

Rob Campbell [00:40:36]:
is my— Not compatible.

Jonathan Bennett [00:40:38]:
That's probably my one hesitation with going to ZFS is just that particular issue. And that's, you know, not necessarily even a technical issue, but—

Rob Campbell [00:40:46]:
Yeah. That's too bad. There's lots of good things about ZFS.

Jonathan Bennett [00:40:49]:
Yeah, I know. Listening to Jeff talk about it, I'm like, Man, maybe I need to, because I'm running an MDADM. I

Rob Campbell [00:40:54]:
don't think you mentioned, like, one of the things I think is the coolest is just the replication, you know, the send, the zsend and zreceive.

Jeff Massie [00:41:04]:
Oh yeah, I didn't get into the heavy duty. This was a

Rob Campbell [00:41:06]:
simple, that's like my favorite

Jeff Massie [00:41:08]:
mirror

Rob Campbell [00:41:09]:
feature, I think, or I think one of the coolest features.

Ken McDonald [00:41:11]:
And this is a headless server, right?

Jeff Massie [00:41:14]:
No, it's, it, well, anymore I kind of treat it like headless, but it does have a GUI.

Rob Campbell [00:41:19]:
Yeah.

Jeff Massie [00:41:19]:
So if I need to get in there, I can.

Rob Campbell [00:41:22]:
Kubuntu. So no point in having a KDE if it's headless.

Ken McDonald [00:41:29]:
I came across a command the other day that would let you make it headless, but still have the GUI there if you ever needed to bring it back up.

Rob Campbell [00:41:37]:
Yeah. Tunnel it over or—

Jonathan Bennett [00:41:40]:
I mean, you could just set it to run level 3.

Rob Campbell [00:41:43]:
Yeah.

Jonathan Bennett [00:41:45]:
That's some ancient wisdom. That's not how you do that anymore, is it?

Rob Campbell [00:41:49]:
Oh yeah, init.

Ken McDonald [00:41:50]:
There's a way to do it through systemd.

Jonathan Bennett [00:41:54]:
Yeah, I bet. I'm sure

Rob Campbell [00:41:55]:
there is.

Jeff Massie [00:41:58]:
Yeah, honestly, the system, I mean, it runs at such low load that, okay, if I'm running KDE in the background, it doesn't matter.

Rob Campbell [00:42:06]:
But that's one thing I kind of miss about the old init. It's just init 0, 1, 2. And it was kind

Jonathan Bennett [00:42:13]:
of fun. Yeah.

Rob Campbell [00:42:14]:
I wanted to say also, Jeff, I thought that was pretty clever, your little hack to steal the drive from the 3-disk array and temporarily use it while you transfer.

Jonathan Bennett [00:42:30]:
Yeah.

Jeff Massie [00:42:30]:
Yeah. Well, and there was a little less—

Rob Campbell [00:42:34]:
Maybe a little risky too, but a little

Jeff Massie [00:42:35]:
bit, yeah. Yeah, but there was a little less risky because I have an LVM of a couple big SSDs on my gaming machine. that also have a copy of the RAID on it as well, which, okay, there's no redundancy there. It's just treated like a big drive, but I have another copy if worst came to worst.

Jonathan Bennett [00:42:58]:
Yeah, because I mean, that is important to note though, that like the most dangerous time for your RAID is while it's rebuilding, because that pushes those drives and you've got For some of that data, you've got one less copy of it than you would have otherwise.

Rob Campbell [00:43:14]:
That, and I imagine having, you know, one disk missing from the RAID while transferring all that data also is riskier for it to fail.

Jonathan Bennett [00:43:26]:
Yep.

Jeff Massie [00:43:27]:
Oh yeah. And that's why I brought up the drive temperature. Mine went up to 50°C. I've got a Fortress 2 case with a big 180mm fan blowing right across the drive cage. So they're, they're getting a lot of cooling air. But yeah, you're stressing your drives for 12 to 24 hours or whatever, however much data you have. You know, it's not a fast process.

Rob Campbell [00:43:48]:
That's, uh, 50°C. That's how hot those, uh, freezers got over there in Europe.

Jeff Massie [00:43:53]:
Yeah, yeah, the drive limit was 60°C on— I've got a couple Western Digitals, well, 2 Western Digitals and then a Seagate. And they, they, they set a drive limit of 60. So I was 10°C away from the thermal limit, but I made sure that, okay, we're staying nice and cool.

Rob Campbell [00:44:13]:
Yeah, those, those hacked freezers were way above their limit, but.

Jonathan Bennett [00:44:16]:
Yeah. Uh, all right. Hey, let's, uh, let's talk, uh, PipeWire. We talked a little bit about this earlier with my OBS story. Ken's got the, the, the rest of the story.

Rob Campbell [00:44:26]:
Now for the rest of the story.

Jonathan Bennett [00:44:28]:
Exactly. What's, what's new in PipeWire, Ken?

Ken McDonald [00:44:31]:
Well, thankfully we can thank Bobby Borisov and Marcus Nestor since they both wrote about the latest PipeWire release. Uh, this time it's 1.6.10. According to Marcus, it fixes a potential crash in the X11 module because error handlers can be called twice, adds support for more output formats by letting the snd file choose the byte order, and improves Bluetooth support by removing some old deprecated functions and fixing an issue with register application calls. According to Bobby, one improvement concerns jack compatibility, where developers added locking around the port array and mixer list to prevent crashes, especially when applications register many audio ports. Another notable fix in PipeWire 1.6.10 eliminates unwanted pops and clicks that could occur when unsigned audio streams finish. The problem was caused by incorrect buffer clearing during stream draining, producing an audible burst at the end of playback. Now I am excited about the GStreamer integration receiving corrections to video plane strides and offsets, plus a fix preventing invalid maximum resolution values from being reported, since this will ensure Pipewire hands a video buffer over to a GStreamer plugin without alignment artifacts. Now, for more details, I am recommending reading Bobbius and Marcus's article or Wim Teijmans' release notes.

Ken McDonald [00:46:19]:
All 3 are linked in our show notes. Yeah, I'm,

Jonathan Bennett [00:46:22]:
I'm looking forward— I was looking at 1.7 and trying to figure out what are some of the things that they're working on there. PipeWire does the even-odd versions where 1.6 is going to be— 1.2, 1.4, 1.6 is their releases. And then 1.7 is kind of their dev versions.

Ken McDonald [00:46:40]:
Ubuntu Studio 26.04 is using 1.6.6 when I last checked. I got to check to see what Amashi's using. Amashi.

Rob Campbell [00:46:52]:
Umashi.

Jonathan Bennett [00:46:54]:
Umashi.

Ken McDonald [00:46:56]:
That's like with Calibre. It's so, it was so easy in the early days to say Calibre.

Rob Campbell [00:47:02]:
Yeah, they messed that one up. But anyway.

Jonathan Bennett [00:47:05]:
I mean, hey, they're the ones that spelled it wrong. Don't look at us.

Rob Campbell [00:47:08]:
I blame them for that. Well, thank you.

Ken McDonald [00:47:09]:
What do you mean? It's spelled proper according to the United Kingdom English.

Rob Campbell [00:47:13]:
Umashi is pretty spelled wrong too. What's that R doing in there?

Jonathan Bennett [00:47:18]:
It's true.

Rob Campbell [00:47:19]:
Arch is in the name. I know that has something to do with it, but I don't know.

Ken McDonald [00:47:23]:
But let's save talking more about Amashi for later.

Jonathan Bennett [00:47:28]:
Oh, for sure.

Ken McDonald [00:47:28]:
What I want to hear about is this proposed file system.

Jonathan Bennett [00:47:33]:
Actually, before we talk about that, I have a question for you, Ken. At what point do you think we're going to look at PipeWire and say it's finished? It is feature complete. We don't need to keep adding stuff onto it.

Ken McDonald [00:47:45]:
I would say some people already say it's feature complete that don't really push the edge of it that much.

Jonathan Bennett [00:47:54]:
Yeah.

Ken McDonald [00:47:55]:
For me, I don't think it'll ever be feature complete.

Jonathan Bennett [00:47:58]:
I mean,

Rob Campbell [00:47:59]:
it's feature complete. Something else is going to replace it.

Jonathan Bennett [00:48:02]:
Probably. You know, I mean, are we in a couple of years, are we going to be streaming 3D? I don't even know what that is.

Ken McDonald [00:48:09]:
Spatial audio?

Jonathan Bennett [00:48:11]:
I was thinking video, like point clouds of 3D video or whatever through it.

Rob Campbell [00:48:17]:
VR.

Jonathan Bennett [00:48:18]:
Yeah, for VR. I don't know. You know, there's always the next step, the next horizon of technology.

Rob Campbell [00:48:26]:
High new standard. Put it up to each lens and it'll do the shifting and

Jonathan Bennett [00:48:30]:
stereoscopic vision. Yep. All right. Anyway.

Ken McDonald [00:48:34]:
We're doing stuffing like this?

Jonathan Bennett [00:48:37]:
No. I bet you that already uses PipeWire to some extent. Anyway, Rob, yeah, let's talk file systems again and fixing corruption.

Rob Campbell [00:48:48]:
Yes. So I thought this was pretty interesting, which is obviously why I'm bringing it to the show to share it all with you.

Jonathan Bennett [00:48:55]:
It's not a high bar, apparently.

Rob Campbell [00:48:59]:
Anyway. We've got a new file system being proposed for the Linux kernel, and this one has a pretty interesting feature. As Jonathan already implied, it can actually detect and repair corrupt data without needing a second copy. The file system is called BeamFS. That's B-E-A-M-F-S. And it uses something called Reed-Solomon error correction technology that's been around for decades and is used in everything from CDs and DVDs to communication systems. Now, file systems like Btrfs and ZFS can already detect data corruption using checksums, but to actually repair that corrupt— corruption, they generally need another good copy of the data. BeamFS takes a different approach by storing additional error correction information alongside the data itself.

Rob Campbell [00:49:56]:
That allows it to repair certain types of corruption, even when there's only one copy available. The project was originally designed for specialized environments like satellites and medical equipment, where radiation can cause individual bits of data to unexpectedly change. And in early testing, BeamFS successfully recovered corrupt data in nearly Every test. Meanwhile, ext3 and ext4 sometimes return corrupt data without detecting it, and BTRFS detected the corruption but couldn't repair it. Now, before you start formatting your drives with BMFS, you know, I don't want Jeff to, you know, start to, you know, wipe out his ZFS drive already since he just finished doing that one. This is still highly experimental. It doesn't support journaling and the developer warns it could cause data loss or even kernel crashes. Kind of ironic for something that's supposed to prevent data loss, but it's early.

Rob Campbell [00:50:58]:
It's early. So I really try to keep AI out of all my stories this week. Jonathan already ruined that one, but I guess that wasn't my story. And honestly, I didn't even realize this story had some AI in it. But by the time I did realize it, I was much too deep into the story to turn back and find something else. So I really tried, guys, but I have to tell you, just full transparency so you know, the developer used Anthropic's Claude AI to help write the file system code and prepare the kernel patches. For some of you, that's— Wah. But that's all I'm going to say about that.

Rob Campbell [00:51:44]:
You know, I thought you should be aware. But anyway, for now, BMFS is just a proposal and hasn't been accepted in the Linux kernel. You don't want an AIF file system. When it comes out, don't use it. But I think it's an interesting concept and it raises 2 questions. Could this approach eventually make Linux storage even more resilient? And 2, perhaps the one some of you out there are asking, Would you really trust a file system developed with the assistance of AI? I would. Hey, that's me. All right.

Jonathan Bennett [00:52:18]:
I mean,

Rob Campbell [00:52:18]:
not the AI.

Jonathan Bennett [00:52:20]:
I don't know. I think we're rapidly approaching the point where we ask the question, would you trust this thing if it wasn't using AI to find the bugs and fix them? It's good to be— oh.

Jeff Massie [00:52:33]:
Well, really either way, it's going to depend on the data. We're just going to have to put You know, hundreds of thousands of drive hours on it and just see, see what the results are. I mean, as Harold Finch says in the, in the, on the Discord, he said, I used to use, uh, RiserFS for quite some time, and that one went away for, for, you know, maybe reasons.

Jonathan Bennett [00:52:55]:
But

Ken McDonald [00:52:55]:
this version, the

Jeff Massie [00:52:56]:
biggest thing I

Ken McDonald [00:52:57]:
see, I

Jonathan Bennett [00:52:57]:
think that all, it's all, it's all pretty much gone away. But go ahead, Jeff.

Jeff Massie [00:53:01]:
Yeah, but I mean, the biggest thing is going to be, is it going to stick around? Is it going to prove itself? Is it going to— because I don't use file systems personally that haven't been around for a long time. I don't want to beta file. I want

Jonathan Bennett [00:53:14]:
the,

Jeff Massie [00:53:14]:
look, this just came out. Yeah, I'll look at it in a few years and see what it says.

Jonathan Bennett [00:53:19]:
Yeah, either that or you've got your play computer over there that you test that stuff out on.

Jeff Massie [00:53:24]:
Yeah. Yeah.

Rob Campbell [00:53:25]:
There's the other consideration is, Yeah, you use that as a tool, as a layer to help protect data corruption. But that's only part of it. If you don't have backups still somewhere else, you're doing it wrong anyway.

Jonathan Bennett [00:53:39]:
Yes.

Jeff Massie [00:53:39]:
Right.

Ken McDonald [00:53:40]:
Well, here's the question. Where are you going to keep your backups when you're in space?

Jonathan Bennett [00:53:48]:
Huh.

Jeff Massie [00:53:48]:
When you're in space?

Rob Campbell [00:53:50]:
Okay, what kind of—

Ken McDonald [00:53:51]:
This is designed to help prevent Corrupted data on your hard— on your storage devices, primarily if it gets hit by radiation. You're going to have more radiation in space than you do at ground level.

Rob Campbell [00:54:08]:
Well, you know how like we have these shielding, you know how we have these little bots of what are they called drones on like Mars and we can communicate with them? Well, other data like backups could do the same thing.

Jonathan Bennett [00:54:22]:
I, I was actually going to say you keep the data, you keep your backup data on the other satellites that are up there with you. Yeah. If you're doing something like Starlink or, um, what—

Rob Campbell [00:54:32]:
Plenty of options.

Jonathan Bennett [00:54:33]:
What's— what is SpaceX calling their, their next generation AI satellite things that they want to launch?

Rob Campbell [00:54:39]:
XAI.

Jonathan Bennett [00:54:40]:
Yeah, yeah, something like that.

Jeff Massie [00:54:42]:
Well, I'm sure they'll do—

Jonathan Bennett [00:54:43]:
they'll do backups between them.

Jeff Massie [00:54:45]:
Or you keep it deep in the bottom of gold mines that are, say, 10,000 feet deep so they're not getting bombarded by all that radiation and

Rob Campbell [00:54:53]:
Gold mines in space.

Jeff Massie [00:54:59]:
On the planet.

Jonathan Bennett [00:55:03]:
All right. Well, let's, let's take a quick break and recover from that. And then we're going to talk about a kernel release. Don't go anywhere. We'll be right back. All right, Jeff, let us know what is up in the 7.3 kernel.

Jeff Massie [00:55:17]:
Well, Yep, 7.3 is out. So Linus Torvalds released it on Sunday, October 18th, right on schedule. And for Linus, on schedule was a small victory this time. The week before release, candidate 7 showed up about 6 hours late because Linus had been traveling to the Kernel Maintenance Summit and Linux Plumbers conference. He lost track of what day it was. You know, but he held tight now on this one. So in the final announcement, he wrote, quote, this week I somehow kept my wits together long enough to keep track of the day of the week. So small victory for Linus there.

Jeff Massie [00:56:00]:
Now with the 7.3 kernel, the headline feature this time is for games, or I think it is anyway. And it applies for people with kind of mid-range and lower-end graphics cards. And it comes from Natalie Vock at Valve, and it's about VRAM, which for those who don't know, that's the super fast memory in your GPU. And here's, here's what it fixes. Now on AMD cards, the kernel hands out VRAM first come, first serve. So just picture a normal evening, you're playing on your computer and you got a browser open, and then you got Discord because you want to Talk on the, uh, Twit Discord and see what's going on. Uh, maybe you got a game launcher or 2 up and each one of those parks some of their data in VRAM. Well, now you're ready to play a game and the game that you, you know, really care about and really want it to run fast.

Jeff Massie [00:56:59]:
Well, it showed up last. So if VRAM is tight, part of that gets pushed down into regular system RAM. Now the GPU can still reach system RAM, but it has to go across the PCIe bus, which is much slower, way slower than having it resident in your VRAM. And this is— you get the result of stutter and lower frame rates on hardware that should be able to handle the game. You might even be thinking like, why is this not working? It should just be fine. Well, in the Linux 7.3 kernel, It can now evict background data that isn't protected out of VRAM to make room for higher priority programs like a game asking for protected memory. So in other words, the game gets the fast memory, then the browser tabs and launchers and whatnot, they get moved to the slower RAM, which you really don't care about for those cases. Now, when they were testing this, they used Indiana Jones and the Great Circle, which is— for those that don't know— is famously hungry for video memory.

Jeff Massie [00:58:06]:
It's kind of a memory pig. With settings pushing it towards the 9 gigabytes of VRAM, the game still ran well on an 8 gigabyte card. Now, that matters because a lot of people are still gaming on these 8 gig cards, and You know, a lot of newer games are not kind to those 8 gig cards. Now, Vox's work is only part of a bigger plan, and more VRAM management changes are still coming in future kernels. So this, this isn't the end of it, but it's a pretty cool first step. Now, there's also a second gaming-friendly change in the scheduler. Linux 7.3 changes how CPU time is divided between groups of processes. called cgroups, which also is how systemd organizes everything on our desktops.

Jeff Massie [00:58:58]:
There's now a new scheduler mode. It defaults called Concur. And the goal is to keep the system responsive when background jobs are fighting for the CPU. Now, the scheduler maintainer, Peter Zidra, tested a game with 8 heavy background processes running. And the minimum frame rate, now this is your lows, went from 3.8 frames per second to 20.6. And the average frame rate went from 48 to just over 57. Now the patch notes say they picked the new default in hopes that it, quote, serves more people than it hurts, end quote. You know that.

Jeff Massie [00:59:41]:
Honestly, that might be the most honest sentence ever written about a default setting ever. But just so you know, if you run it and you're like, oh, this is trash, it's terrible, you can switch the old behavior back on. So it's still— it's not a permanent setting, or it's not a setting you can't reverse if it's, for whatever reason, your system doesn't like it. Now, we did get some file system love too. ext4 gets faster parallel direct writes, a faster Fallocate, which what that does is it allocates disk blocks instantly without performing data I/O. So basically, instead of saying, okay, we're going to clear all the data and then we're going to allocate this mem— or this storage space, it just kind of moves the pointers. It's not a secure, uh, erase or anything like that. It just says, okay, here's the new boundaries and away you go.

Jeff Massie [01:00:40]:
It also has a fix for a buffered write slowdown that was occurring. Small direct writes got faster on both ext4 and XFS. And a ByteDance engineer tested it on a PCIe Gen 5 NVMe drive and saw 4K random reads go from about 1.92 million operations per second to 2.19 million, with bigger gains when you're using io_uring, which we've talked about in the past. And for Btrfs fsync, the call that makes sure your data actually hits the disk can be up to 5 times faster in some cases now. So a lot of file systems got some love. On the hardware side, there's a driver for Valve's new 2026 Steam Controller. Valve also added a driver for the MSI Claw. Built-in gamepad.

Jeff Massie [01:01:35]:
Um, you know, Apple— this will be one for Rob— Apple Silicon fans get early support for the M3 Pro, the Max, and the Ultra. The M3 Max and M3 Ultra. Early is the key word here though, because it's not nearly as usable as the M1 and M2 support. So don't, don't think things are going to be great. It's still very early and things are rough. Uh, Novu got the open NVIDIA driver gains— got the open NVIDIA driver gains video decode support, which NVK needs for Vulkan video. And there's a nice— here's a nice little small one. The kernel now checks what a USB-C cable can actually do before trying to run DisplayPort or Thunderbolt on it.

Jeff Massie [01:02:23]:
So if you got that random cable laying around and you're not sure it can do it or not, Your kernel's going to check for you. Now, one question that comes with this is whether or not 7.3 becomes this year's long-term support kernel. Some coverage, you know, some articles are saying, oh yeah, yeah, this is our LTS for this year. But Greg Crow-Hartman usually picks the last kernel of the year and Linux 7.4, 7.4 is due out right around Christmas. So 7.4 might take that title instead. And we're not going to know which one it is until Greg says so. Speaking of 7.4, just so everybody's aware, with the 7.3 out, the 7.4 merge window is now open. So take a look at the link of the article in the show notes for more details and Happy upgrading.

Jonathan Bennett [01:03:22]:
And 7.4 is really intriguing, by the way.

Jeff Massie [01:03:26]:
Yeah, I thought about doing some stuff on it. I thought, well, you know what, I'm just gonna maybe wait till the merge window closes because sometimes stuff that gets pulled in gets pulled out before the merge window closes.

Jonathan Bennett [01:03:38]:
We shall hope that does not happen here because there's some really cool stuff that got pulled in, mainly the The AMD HDMI 2.1 support is finally in 7.4. That got pulled in. So I guess you're right. We wait for the end of the merge window, make sure things aren't too terribly broken. But yeah, here in the next couple of weeks, I'm sure I'm going to go grab and start playing with that 7.4 kernel way ahead of time.

Jeff Massie [01:04:08]:
And it will be a great Christmas present for those that don't play with the RCs.

Jonathan Bennett [01:04:11]:
Absolutely.

Jeff Massie [01:04:13]:
If it stays on schedule.

Ken McDonald [01:04:15]:
So which one are you betting on for the next long term, 7.3 or 7.4?

Jeff Massie [01:04:23]:
I— my personal bet is I bet it's going to be 7.3 because 7.4 is going to come so close to the wire that Greg will just play it safe and say, okay, 7.3 is what it is.

Jonathan Bennett [01:04:33]:
I, I'm gonna say 7.4.

Rob Campbell [01:04:37]:
What about you, Ken?

Ken McDonald [01:04:39]:
I'm gonna go with 7-3, uh, because there's still the chance that either one of them could end up having to go an extra week before they get merged.

Jonathan Bennett [01:04:52]:
Rob, what do you think?

Rob Campbell [01:04:53]:
I'm gonna go with 7-4. I'll go with Jonathan because he's the smartest one here. You shouldn't—

Jeff Massie [01:05:02]:
for those that are only listening on audio, we almost had a spit Oh yeah,

Jonathan Bennett [01:05:10]:
my tea. I was drinking my tea and he said that. Oh my goodness. Warn me before you do that, man. All right. Ken, you've been playing with Umachi, haven't you?

Ken McDonald [01:05:20]:
Yes, I have.

Jonathan Bennett [01:05:21]:
What do you think of it?

Ken McDonald [01:05:23]:
I like it. Pretty much it allows me to stay on the keyboard the whole time. Though muscle memory still has me reaching for that mouse. But I'm going to give you some updates on Amarsi that Bobby Borisov and George Whitaker wrote about. They talked about the release of Amachi. Did I get it right, Rob?

Rob Campbell [01:05:48]:
You got it right. Just type it down, you know. Amachi

Ken McDonald [01:05:53]:
4.0.5 that was released on the 18th of October. Focusing on refining the existing desktop experience while preparing the distribution for its next major feature release. Now, according to Bobby, Umachi 4.0.5 updates the default kernel to Linux 7.2.8 and features the BOR, that's Burst Oriented Response Enhancer, scheduler to improve responsiveness under desktop workloads. The distribution's Arch packages have also been refreshed to the October snapshot, including Qt 6.12. Now, application-related improvements include Steam installation now selecting the appropriate graphics driver. EasyFX 8 gets fixes for audio device handling and filters. And the Windows virtual machines no longer fail to launch because of directory permission issues. According to George, one of the most significant changes in Umachi 4.0.5 concerns laptops equipped with hybrid NVIDIA graphics configurations.

Ken McDonald [01:07:16]:
Umachi optimizes the environment variables used by hybrid NVIDIA NVIDIA laptops, improving how the distribution configures graphics-related behavior on these systems. This is particularly relevant for Wayland environments where graphics device selection and compatibility can depend on how the compositor, graphics drivers, and individual applications interact. Now, Thursday, I personally updated Umachi to 4.0.6. After some research, I found that Ryan Hughes published this version on the 19th, just a day after 4.0.5. Now, its release notes state Umachi 4.0.6 resolves a few small hiccups specific users encountered with version 4.0.5. Apparently, machines with OpenCLaw installed could not get past a migration on 4.0.5. Um, do apologize if, um, I'm just going to say DHH, that's easier. Also provided a securely related fix for the screen where it's sometimes not locking before suspend.

Ken McDonald [01:08:33]:
As always, to keep it brief, I've only touched on what I thought important. More details are available from both Bobby George and Umarch Central via the links in our show notes. And I've got a special treat for y'all when we get to the end of the show.

Rob Campbell [01:08:52]:
I think that OpenClaw update issue was fixed, but there's also some other really interesting stuff with Umarch in the news this week that you didn't touch on because you mostly were talking about that release, but they have their first ARM installer. One ISO targeting multiple ARM platforms, forms. Uh,

Ken McDonald [01:09:12]:
can I put it on a Raspberry Pi 5?

Rob Campbell [01:09:16]:
Probably. NVIDIA N1X support, uh, Snapdragon X support. And then the Umachi 4.5 RC1, Release Candidate 1, has a major, uh, is it, has major new desktop features. Better window management, notifications, and support for 50 languages. So those are some of the headlines that I saw this week and looked at. And the ARM stuff is really interesting to me.

Jonathan Bennett [01:09:47]:
It looks like, yes, you can put it on a Raspberry Pi 5. Apparently really fast too. At least one article claims you can do it in 38 seconds.

Rob Campbell [01:09:57]:
Hmm. Yeah, I remember when, Oh, I guess I don't remember when Umachi first came out, but when I first started hearing about and looking at it, one of the big goals was fast install.

Jonathan Bennett [01:10:08]:
And

Rob Campbell [01:10:09]:
I've never seen it anywhere near as fast as, uh, some of the claims. And not even the videos that I've seen on it were as fast as it claims, but, uh, it's still pretty quick.

Ken McDonald [01:10:19]:
What's the claims?

Rob Campbell [01:10:21]:
Oh, I think it's like 30 seconds.

Ken McDonald [01:10:25]:
Install?

Rob Campbell [01:10:26]:
I want to say maybe 2 minutes. It was very low.

Ken McDonald [01:10:31]:
I'll admit, compared to other systems, it did seem to be faster to install, but customizing it's where it takes time.

Jonathan Bennett [01:10:40]:
Yeah, well, I mean, that's true of all of these sort of, how do I want to put it? Like I've heard them called ricing desktops.

Rob Campbell [01:10:50]:
You can always choose not to customize it.

Jonathan Bennett [01:10:52]:
Yeah,

Rob Campbell [01:10:52]:
with the default

Jonathan Bennett [01:10:54]:
intended to be customized, intended to be, you know, put spinners and neons and all that good stuff on it.

Rob Campbell [01:11:00]:
Yeah, my install is not all that customized really. I played with the customization at the beginning, but now I'm pretty much just running the default.

Ken McDonald [01:11:07]:
My customization is more on being able to basically have the exact same data available seamlessly When I'm dual booting from either one.

Jonathan Bennett [01:11:21]:
That makes sense. I can see that.

Jeff Massie [01:11:23]:
I have dark mode and a wallpaper.

Jonathan Bennett [01:11:27]:
I mean, what more do you really need?

Jeff Massie [01:11:29]:
Yeah. I

Ken McDonald [01:11:30]:
don't even turn dark mode on. I think it

Rob Campbell [01:11:34]:
just defaults, but dark.

Jonathan Bennett [01:11:37]:
So many things have defaulted to dark mode now. I tell you what I wish had a decent dark mode, and maybe there's an add-on to do it, but Google Docs. Google Docs is just blinding. I bet there's some add-on to get that.

Ken McDonald [01:11:50]:
But it makes such a great light for when you're doing a podcast.

Rob Campbell [01:11:54]:
Yeah, you don't need a ring light.

Jeff Massie [01:11:57]:
Yeah, you can see the reflection in my glasses and the glass back there.

Ken McDonald [01:12:03]:
And above your glasses?

Rob Campbell [01:12:05]:
And your head.

Jeff Massie [01:12:08]:
No, that's from my light up there and my light up there. Straight

Rob Campbell [01:12:12]:
up. But I can read the text on your forehead there. Come

Jonathan Bennett [01:12:16]:
on.

Rob Campbell [01:12:17]:
Yeah.

Jonathan Bennett [01:12:20]:
All right.

Jeff Massie [01:12:21]:
Let's just— It's skincare.

Jonathan Bennett [01:12:22]:
Yeah. I think it's actually time.

Ken McDonald [01:12:24]:
I think Rob needs a break to grab a bite to eat.

Jonathan Bennett [01:12:27]:
Yeah, it's time for a break. We're not going to let him go eat. He's not diabetic or anything, but we're going to take a quick break. And then we've got some last few stories coming up. Don't go anywhere. They're good ones. We'll be right back.

Rob Campbell [01:12:42]:
All right. So it's, it's been a while since I've had a here's something Jonathan should buy segment. But I think I found something. And Jonathan, this one's really cheap. Well, sort of. Brax Technologies announced the Brax Winx, a small privacy-focused Android phone that you can reserve today for only $30. All right, now before you get too excited, that's just the deposit. The actual phone costs $550 or $449 if you pay in full during the first pre-order batch.

Rob Campbell [01:13:16]:
But it's still a pretty interesting little device. For starters, it has a 4-inch OLED screen. Um, for those watching, it's pretty much that thing right there. Um, uh, 8 gigs of RAM. 128 gigs of storage and a 50-megapixel camera. I think that RAM alone is probably where most of that cost is. If you lower that RAM amount, the entire phone probably could be around $30. Anyway, if it also has something we haven't seen much lately, and that is a physical QWERTY keyboard that attaches To the phone.

Rob Campbell [01:13:58]:
So if you've been— it's attached to the phone. So if you've been missing your old BlackBerry, this might be right up your alley. But the real selling point is privacy. The Wink runs BraxOS, a de-Googled version of Android that doesn't include Google Play Services or require a Google account. You could choose between a full Android experience if you want, or the stripped-down mode focused on messaging. calls, and navigation. There's also a physical privacy switch that can disable things like the cellular modem, microphone, and camera. And for those who are really paranoid, the Linux type, you know, the phone supports automatic data wipe after too many failed unlock attempts, or, uh, when, uh, or it wipes when a special duress password is entered.

Rob Campbell [01:14:48]:
Though I think we've talked about, talked on the show about some potential legal issues with the duress password and how it could actually get you in trouble under certain circumstances. So just be careful where you use that, I guess. But Brax is promising software updates through 2032, which I guess isn't too bad, but I don't know. It's about 5-ish years. You know, the phone isn't shipping until 2027, uh, with the first batch expected July. So it's barely 5-ish years, but there are some obvious downsides. Without Google Play Services, certain Android apps, like especially banking apps, might not work. So Jonathan, for just $30 down, you could reserve your de-Googled Android phone.

Rob Campbell [01:15:41]:
with a physical keyboard and hardware privacy controls today, and then pay the rest of it later. You can check it out at braxtech.net. That's B-R-A-X-T-E-C-H.net.

Jonathan Bennett [01:15:55]:
2 primary problems. One, if I'm going to go the de-Googled route, I would want to run probably LineageOS as kind of the, as far as I can tell, the best, like most open source solution for doing this. And 2, The problem with de-Googling a phone is so much of the good stuff is from Google. It's like, I have a phone. Now what do I do with it? Can't check my email. Call and text? Who does that with their phones anymore?

Rob Campbell [01:16:25]:
I mean, all I use my phone for is to text. I don't really use it for call. I use it for ignoring phone calls.

Jonathan Bennett [01:16:30]:
I ignore phone calls and I get spammy text messages.

Rob Campbell [01:16:34]:
Yeah, I guess maybe text messages too. But web browser, that's about it.

Ken McDonald [01:16:39]:
I use my phone all the time for calling or answering calls from my wife.

Rob Campbell [01:16:43]:
Yeah, you would.

Jonathan Bennett [01:16:44]:
Yeah, Ken, you're from the previous generation. That checks out.

Rob Campbell [01:16:48]:
Okay, boomer.

Jeff Massie [01:16:51]:
I got to be right in there with Ken. I got to back him up on this one. That's— I— but having the nice contact list in there, if you're not on my contact list, I'm not answering the phone.

Jonathan Bennett [01:17:03]:
Yeah. Oh, yeah, for sure. I, for the longest time, I had to answer the phone whether I knew the number or not because I ran my own IT business and that's how I made money is picking up the phone and coming and helping people. I do not miss that. My phone's on silent almost all the time now. Check in at the end of the day.

Ken McDonald [01:17:21]:
I only use access to my second line with a de-Googled phone.

Jonathan Bennett [01:17:28]:
Because Google Voice.

Ken McDonald [01:17:29]:
Yes.

Jonathan Bennett [01:17:30]:
Can you still do SIP? Can you still do, uh, talk to Google Voice with SIP? I know for a while you could.

Ken McDonald [01:17:35]:
I haven't tried it.

Rob Campbell [01:17:36]:
Yeah,

Jonathan Bennett [01:17:38]:
so you might be able to, to go get a, a, um, you know, a generic SIP client and talk to it that way.

Rob Campbell [01:17:46]:
So I've had Google Voice set up with a, a PBX, like some Asterisk,

Jonathan Bennett [01:17:52]:
uh,

Rob Campbell [01:17:53]:
PBX before. Yeah.

Jonathan Bennett [01:17:54]:
The

Rob Campbell [01:17:55]:
other

Jeff Massie [01:17:55]:
thing is if you're not very popular, nobody calls you.

Jonathan Bennett [01:17:59]:
You would think. But your number escapes because you put it on some form. And then the next thing you know, Sanders wants you to help one more time.

Rob Campbell [01:18:09]:
I don't understand.

Ken McDonald [01:18:11]:
Which is why I make sure it's the Google Voice number that gets on those forms.

Rob Campbell [01:18:17]:
Like my phone number isn't publicized Anywhere that I can think of, yet people—

Ken McDonald [01:18:24]:
You haven't looked at your LinkedIn.

Rob Campbell [01:18:26]:
Do I have my phone number on there? Do I?

Ken McDonald [01:18:30]:
Go look.

Jonathan Bennett [01:18:31]:
I mean, there's a limited number of phone numbers, like in the US, there's a limited number of phone numbers, and I'm convinced that they just call all of them.

Rob Campbell [01:18:38]:
Well, they seem to know my name too, so.

Ken McDonald [01:18:42]:
It's a real easy script to write to call all those phone numbers for the US.

Rob Campbell [01:18:46]:
Yeah, but they know my name when they call.

Jonathan Bennett [01:18:48]:
Yeah, so your data got leaked somewhere. Do you punch yourself into Have I Been Pwned from time to time? Are you on their mailing list?

Rob Campbell [01:18:57]:
It's been a while, but—

Jonathan Bennett [01:18:59]:
I bet you they'd have the details.

Rob Campbell [01:19:02]:
It could be released in a breach somewhere like that, and then don't.

Ken McDonald [01:19:06]:
Which breach?

Jonathan Bennett [01:19:08]:
Yes.

Ken McDonald [01:19:08]:
That's the question.

Jonathan Bennett [01:19:10]:
Yes.

Jeff Massie [01:19:10]:
Yeah, like I said, I keep— Delete me going all the time. So it helps. I watch all that stuff and put up the

Jonathan Bennett [01:19:20]:
mute.

Jeff Massie [01:19:21]:
And if you get spam calls or whatever, you never answer. Never. I hear people go, oh, I'm going to play with them. All you're doing is verifying it's a live number.

Jonathan Bennett [01:19:30]:
Yeah. Although there are some fun open source projects where basically you run an Asterisk server that I don't think, I don't know if they're using actual AI, but the one that I saw was, it was just a series of prerecorded messages of some old guy being like, oh yeah, well, you know.

Rob Campbell [01:19:46]:
I used to have an app for my phone that would do that for like, obviously spam numbers. And you could have a whole bunch of things. Like some were like celebrities just saying dumb things, like prerecorded stuff. Like, hey, oh, I can't hear you.

Jonathan Bennett [01:20:02]:
Yes.

Ken McDonald [01:20:04]:
How many of y'all are able to access your phone information From a browser?

Jonathan Bennett [01:20:11]:
Uh, some of it, yeah.

Rob Campbell [01:20:12]:
Yeah, some of it, sure.

Jonathan Bennett [01:20:16]:
What phone information are you thinking of?

Ken McDonald [01:20:19]:
I use— I, I'm able to send text messages via browser to other people.

Jonathan Bennett [01:20:26]:
Yeah, there's a few things that let you do that. Uh, KDE has a plugin that lets you do that that I ran for a while. I don't think I've set it up with the new phone. It wasn't great. Like, it worked, but it was not the best experience.

Ken McDonald [01:20:40]:
It wasn't GPU dependent, was it?

Jonathan Bennett [01:20:47]:
No. I think Ken is really stretching for a segue here.

Ken McDonald [01:20:50]:
Yes, I am.

Jonathan Bennett [01:20:51]:
Really stretching.

Jeff Massie [01:20:54]:
But he got the letters wrong.

Jonathan Bennett [01:20:57]:
CPU. Yeah, I was going to say. Ryzen CPU.

Rob Campbell [01:20:59]:
That is such a bad segue. Bad attempt at a segue. I think you lose your next round.

Jonathan Bennett [01:21:07]:
You mean it's not like B-grade movies where it comes around again and it's so bad it's good?

Rob Campbell [01:21:12]:
No, that is never going to—

Ken McDonald [01:21:15]:
That's the Toxic Avenger there.

Jonathan Bennett [01:21:18]:
All right, Jeff.

Jeff Massie [01:21:20]:
So, you know, and I've talked about this in the past on the show. I've done some overclocking, benchmarking, I've compiled kernels for performance, you know, shared the information here with everyone. And it's gotten a little harder than it used to be. But there's also tools that make it easier. So if you've ever wanted to have some time on your Ryzen CPU to overclock on Linux, you know, it got a little bit easier. Now, for example, on Windows, AMD gives you Ryzen Master. It's an official app with sliders and buttons. And, you know, it kind of overclocks based on the silicon lottery, i.e., how good your cores actually are, and kind of makes overclocking very simple.

Jeff Massie [01:22:13]:
On Linux, you get 2 choices. One, you reboot and you go into the BIOS and you dig through menus and settings and You're manually playing with clocks and clock multipliers and voltages and all sorts of timings and trying to get things right. Or there are community tools that talk to the processor's management firmware directly, but they're using interfaces that AMD never officially documented for the job. And they work most of the time. And the people that build them, they do great work. So I'm not down on that. But you're always one firmware update away from something breaking because they were never meant to be an official API. So who knows what's going to change in the future? Well, things seem to be changing.

Jeff Massie [01:23:04]:
AMD engineer Shyam Sundar SK has posted kernel patches for a new driver called AOD, which is short for AMD Overclocking. Hardware Busters calls it AMD's, and this is a quote, first official Ryzen overclocking driver for Linux. Well, what does it actually let you change? Now, this is just the first version, so more to come. But first thing is it handles 3 power limits. And if you've ever poked at Precision Boost Overdrive, These are going to sound familiar. So PPT, or package power tracking, is the maximum power the whole CPU package is allowed to pull in watts. You can change TDC, or thermal design current, which is how much current the chip can draw for sustained loads, and EDC, or electrical design current, which is the short-term peak current for quick spikes. These are just your little Surge and then come down.

Jeff Massie [01:24:11]:
Uh, you can, you can rate— you can raise those limits and the CPU can boost harder for longer if your cooling can keep up. A lot of times with modern CPUs, we're dealing with the thermal limits more than the actual silicon limits as far as voltage and clocking and things like that. It just generates too much heat. So you You're going to want to make sure you've got really good cooling if you really want to make a difference. Otherwise, you're probably not going to really see anything. Conversely, if you lower them, you get a cooler, quieter, more efficient machine. And, you know, the second one is honestly something a lot of times people care about. You know, not everybody wants more speed, and some just want their fans to calm down.

Jeff Massie [01:24:58]:
They don't want to sit next to a screaming machine no matter how fast it is. Now, the driver also reports the range of values your platform allows for each limit and the step size. So the tool tells you what you can do and what's legal before you try something silly. So if you go, oh, I want to make my CPU handle 6 times what it wants to do, it's going to go, Whoa, that's not an approved setting. It can show the current memory clock, but it's read-only for now. And it can turn simultaneous multithreading or SMT on or off at the firmware level. Now that's different from the no SMT kernel boost option. Now we're talking in BIOS where the threads still exist and Linux just ignores them.

Jeff Massie [01:25:58]:
With AOD, the platform itself turns it off. Now, here's the catch. None of this happens live. So you read the allowed range, you write a value, now you reboot. And the settings are stored in firmware and they take effect. On the reboot. So it's, it's not going to be a turn the dials and see instantaneously what happens. There's still some work in it, but it's, it's taking out some of the trudging through deep menus in BIOS.

Jeff Massie [01:26:34]:
Now the firmware currently doesn't even report back what the active limits are, so the settings that you write are write-only. You just tell it what you want and you trust it. So if you're thinking, I don't think it took it, you're just going to have to rewrite it and hope, you know, that that second time it took it, or maybe the results you're getting are with the new values. You can't read those values out. Under the hood, the driver uses the WMI, which is a firmware interface that a lot of laptop vendors also use. It exposes everything through SysFS and, you know, the files under /sfs. So true Linux fashion, overclocking your CPU starts out as echoing a number into a file and rebooting, which honestly is the most Linux way possible to do it. I mean, you know, and I mean that with love.

Jonathan Bennett [01:27:29]:
Absolutely.

Jeff Massie [01:27:31]:
Yeah, but that is so Linux, right?

Jonathan Bennett [01:27:33]:
Yes.

Jeff Massie [01:27:33]:
The firmware supports a lot more commands than this first version uses. And the patches on the interface say can grow as needed. There's also a raw debugging interface, but the documentation warns, and it's a quote, it's not a stable API. So as the authors say, tool authors should not build on it. So consider yourself warned, tool authors, i.e., Whatever you write is probably going to change in the future. What about an app? So far, AMD hasn't shipped one. There's no Ryzen Master for Linux in these patches. The expectation is that the community and the desktop environment will build the friendly front ends the same way, you know, community tools grow around AMD's Radeon GPU controls.

Jonathan Bennett [01:28:28]:
Mm-hmm.

Jeff Massie [01:28:29]:
AMD doesn't provide them. They just kind of give you the interface to them. You know, so they give you the official documented interface and it's up to the community to make the rest of it. The big caveat, these patches are under review and they're not merged code. So it's not a done deal yet. This 3-patch series went in the platform driver mailing list on October 10th, and it was not part of Linux 7.3. We're not even sure which kernel will carry it. I mean, the merge window is open, but it still has to get through reviews, and, you know, reviews can lead to changes, and it may or may not make the merge window.

Jeff Massie [01:29:10]:
I wouldn't count on it for this cycle to hit the merge window, but fingers crossed. You know, but this is a meaningful shift. You know, for years, Ryzen tuning on Linux has meant, you know, kind of either borrowing access that AMD never meant to give, or like I said, trudging through the BIOS and hoping you know what all those different acronyms are, looking them up and try to figure out how they interact with each other and how they all interface. You know, so AMD is kind of building the front door to walk in and start overclocking. You know, it's, it's a small step, 3 power limits and a reboot, but it's official. And it can grow. So this is just the, again, the beginning. Take a look at the article linked in the show notes for full details and happy under or overclocking.

Jonathan Bennett [01:30:05]:
So at the very beginning of the story, you said something that as a minor celebrity really resonated with me. And that is, I just want my fans to calm down and stop screaming. It's a pretty good line. It's too long for a show title, but it's a pretty good line. The other thing that really comes to mind is I am so glad that AMD is not making an official tool for this because I've seen their official tools on Windows and they're all pretty terrible. Let the community do it. That's the right way to do it.

Jeff Massie [01:30:39]:
Oh yeah, the community will do an awesome job.

Jonathan Bennett [01:30:42]:
Yes. Absolutely.

Ken McDonald [01:30:47]:
I think I overreached with my segue.

Rob Campbell [01:30:51]:
Yeah.

Jonathan Bennett [01:30:53]:
Yes, I think we came to that conclusion already.

Ken McDonald [01:31:00]:
So now that you're going to have this AMD overclocking driver, are you going to have something to overwatch it?

Jonathan Bennett [01:31:11]:
Well, there's lots of stuff going on that needs overwatch. So there was a, yeah, there's a bit of a story this week on security. In fact, there's about 3 or 4 or 5 different stories on security that we can go over real quick. And it's kind of the week that just keeps on giving. So one of the first big things to talk about is there is a certificate authority, eMudra, which is an Indian CA. They sell certificates. And it turns out that their certificate, the secret key behind it, you know, the thing that's supposed to be like 1,024 bits long, uh, it was 4. Not, not 4 bits long.

Jonathan Bennett [01:32:03]:
In fact, you could do it in 3 bits because the value of that secret was 4. And I have to give a shout out to Randall Munroe on XKCD because he totally called this like a decade ago. I'll throw it in the show notes, but there is an XKCD comic that says essentially the ISO-approved random number is 4. And yeah, it really came back for all of us on this one. It's XKCD 221. And he even has a pseudocode here of int getRandomNumber, return 4. Wow. It was chosen by dice roll.

Jonathan Bennett [01:32:44]:
It's guaranteed to be random. Yeah, so it was really intriguing to see that. And the reason that this is a problem, of course, like if you're not up, if it doesn't immediately jump out to you what the problem is, it is this is such a small number, it can be easily, easily brute-forced. And once you brute force it, you can then sign what appears to be legitimate certificates for any website. And so until those got pulled, which all the browsers pulled them very, very quickly, it was possible to generate malicious certifications. That was going to be the big story of the week, of course. And then PHP happened. If anyone, any sysadmins out there, I'm sure know exactly what was up with this.

Jonathan Bennett [01:33:36]:
The vulnerability was PHP2Shell. And it was a memory corruption vulnerability. And essentially what can happen inside PHP, and I didn't realize this, this is a known problem. You can start a file upload and get the memory into a certain state. and then kill PHP and it starts tearing things down. But because it was still in the process of doing the file upload and like setting its memory up, it can have stale values in its memory structure. And essentially you can get it to then do things, read and write from unallocated memory or memory locations that aren't what it thinks they are. So This was found by LLM.

Jonathan Bennett [01:34:26]:
And it was, well, it was going to be patched, I think yesterday, Thursday or Friday of this week. And before the patch was released, the proof of concept got released. I'm not sure if someone found it again, you know, if it was an independent discovery or if it was just leaked. But this one's pretty serious. And if you run PHP anywhere, you should go and get it patched for sure. I

Rob Campbell [01:34:56]:
think it was actually Wednesday.

Jonathan Bennett [01:34:58]:
Was it Wednesday? Okay. Yeah, sometime this week. And then there's more. There's a Node.js problem. There's an HTTP problem. And then one more that I thought was really interesting is The, uh, the, the problem in Broadcom Wi-Fi. And, uh, this has apparently been in the Wi-Fi drivers since 2010. And the, the really unfortunate thing about this one is when the guy found it and applied for a CVE for it, Broadcom said nope, because they are their own, um, CVE authority.

Jonathan Bennett [01:35:38]:
They closed it. And they're like, no, no, no, that's not a CVE. And so, you know, a whole bunch of these old routers If you have not upgraded to a modern OpenWRT on that router, if it's got Broadcom in it, it's probably vulnerable to this particular bug. So lots of crazy stuff going on this week. Yeah, a bunch. Rob, you follow this stuff too. What stuck out to you?

Rob Campbell [01:36:04]:
Yeah, I mean, I know there was those 2, there are 2 PHP vulnerabilities and Yeah, did a bit of patching right away that week. And I know we were seeing vendors, our site is down right now, or patching, we'll be back up soon. And

Jonathan Bennett [01:36:23]:
yep, a

Rob Campbell [01:36:23]:
lot of that.

Jonathan Bennett [01:36:24]:
Yeah.

Rob Campbell [01:36:27]:
I think

Ken McDonald [01:36:28]:
I may have run into one or two of those sites.

Rob Campbell [01:36:33]:
Yeah. And what Broadcom, I didn't, I didn't, I didn't hear about that one. But Yeah, I mean, that's the— whoever, whoever thought Broadcom was ever going to do anything good?

Jonathan Bennett [01:36:47]:
Well, I wasn't going to be quite that hard on them, but

Rob Campbell [01:36:49]:
I would.

Jeff Massie [01:36:51]:
I just keep

Rob Campbell [01:36:52]:
going back to ruin everything they touch.

Jeff Massie [01:36:56]:
2 is a couple, 3 is a crowd, and 4 is random.

Ken McDonald [01:37:02]:
Then where does 42 fit in?

Jonathan Bennett [01:37:04]:
That's the answer.

Rob Campbell [01:37:05]:
Yeah.

Jeff Massie [01:37:06]:
They didn't have that big a dice.

Jonathan Bennett [01:37:09]:
They didn't have that big a dice.

Jeff Massie [01:37:13]:
It was a d20. I mean, that's the standard.

Jonathan Bennett [01:37:16]:
Yep.

Rob Campbell [01:37:17]:
Yeah. I mean, everyone hates Broadcom's the one everyone hates for VMware. So just remember that.

Jeff Massie [01:37:25]:
More than Cisco?

Ken McDonald [01:37:26]:
Here's a question.

Rob Campbell [01:37:27]:
Oh, definitely more than Cisco. Broadcom is like the worst. In the whole industry. I mean, the only ones who maybe come close is Adobe.

Jeff Massie [01:37:38]:
But Adobe just— that level of hate.

Rob Campbell [01:37:41]:
Yeah, I mean, the thing is Adobe just has their own products and they do their things with them. Broadcom buys a good product and ruins it and then buys another good product and then ruins it.

Ken McDonald [01:37:53]:
What happens when you XOR 4 with any other number?

Jonathan Bennett [01:37:59]:
It depends. Let's see. It's going to be— if you're just doing the 2 bits, it's going to be no change because they're both— or no, no, no, no, that's not true for an XOR. It's going to invert them. It's going to invert those 2 bits because 4 is 11 in binary. Is that where you were going with that, Ken?

Ken McDonald [01:38:26]:
It's 1-1.

Jonathan Bennett [01:38:28]:
Isn't it?

Rob Campbell [01:38:30]:
I think Ken has some weird segue to saying it's time for a break, but I don't know.

Ken McDonald [01:38:35]:
0-1.

Jonathan Bennett [01:38:36]:
Oh, it's 1-0-0. I counted in binary wrong in my head. I did math on camera and it came around to bite me.

Rob Campbell [01:38:45]:
Yeah, it's 1-0-1, right?

Jonathan Bennett [01:38:46]:
Yeah, people are now yelling at me. No, 4 is 100. Yes, 4 is 1-0-0.

Rob Campbell [01:38:51]:
Yeah, one.

Jonathan Bennett [01:38:53]:
I did math on camera. What was I thinking? Anyway, we're going to take a quick break. I'm going to recover from my shame and then we're going to do some tips. Don't go anywhere. We'll be right back. All right. It is time for tips and Ken is going to lead us off. What do you have for us?

Ken McDonald [01:39:10]:
Well, that's all. Since I'd covered Calibre, probably time to start to covering some of the command line tools that get installed as part of Calibre. So I'm starting with the command that allows you to interface with the Calibre Books database. It's called calibre-db. Let me go ahead and bring my screen up here. And is that big enough for everybody to read that's watching the video?

Jonathan Bennett [01:39:40]:
It's not too bad. Yeah,

Rob Campbell [01:39:41]:
works for me.

Jeff Massie [01:39:42]:
Looks good.

Ken McDonald [01:39:48]:
And as I said, it's CalibreDB. That's C-A-L-I-B-R-E-D-B. And, uh, it's got the, uh, often used --help or -h to give you some basic information about it. That's the basic information. For those of y'all listening, I just did hit enter and it gave me a screen and a half worth of information. Basically, the usage for CalibreDB is CalibreDB followed by a command. I'm going to actually call it an action command, then options or arguments that you're going to enter. And it is a command line interface to the Calibre books database.

Ken McDonald [01:40:39]:
Some of the commands you can use would allow you to list your books, add, remove, add the format, remove a format, show the metadata for, even save searches, add custom columns, list your custom columns, even check your library. I'm not going to demonstrate all of those for you. I'm just going to demonstrate 2 of them. The first one, of course, will be the list since that's the one I've used the most because that way from the command line. And, uh, with any of those, uh, commands or action options, you can use the --help after it and it tells you what it will do. Add as probably most of y'all assumed list will list the books available in the Calibre database. What you didn't know is what the options for it are. You can, uh, have options for listing, uh, the field by fields.

Ken McDonald [01:41:44]:
You can do a sort by ascending. You can do searches in your database and set up separators, uh, to separate the fields. You can prefix your file paths, uh, you set up a template, and you can also use, uh, global option, one of which is --library-path= where you tell it where the library is. I'm going to use that one with The library I'm going to look at, it's my paper library. Yes, I've got a paper library documented in Calibre.

Jonathan Bennett [01:42:30]:
Nice.

Ken McDonald [01:42:33]:
It's all the physical books I have. For those of y'all listening, I just typed in— let's scroll back up so I can read it.

Rob Campbell [01:42:43]:
So it's like your card catalog?

Jonathan Bennett [01:42:44]:
Yeah.

Ken McDonald [01:42:45]:
Yep.

Jonathan Bennett [01:42:46]:
That's actually cool.

Ken McDonald [01:42:47]:
I typed in calibre db list --library-path=, and here I use quotes because I have a space in it, uh, /home/dad/public/calibre libraries/paper/. I've gotten in the habit I blame rsync for this, which is probably good.

Rob Campbell [01:43:17]:
Can you use the Dewey Decimal System for the ID?

Ken McDonald [01:43:22]:
The ID is automatically generated by Calibre when you enter it.

Jonathan Bennett [01:43:25]:
Most of these are fiction titles too, so Dewey Decimal doesn't really apply there.

Ken McDonald [01:43:33]:
Yeah, Star Trek IV: The Final Frontier.

Jonathan Bennett [01:43:36]:
Yep. Lots of Star Trek books. I recognize several of these titles.

Ken McDonald [01:43:40]:
And you probably recognize some of these.

Jonathan Bennett [01:43:42]:
Hardy Boys.

Ken McDonald [01:43:44]:
Yeah.

Jonathan Bennett [01:43:45]:
Yup. How many books are in the library? How many do you have?

Ken McDonald [01:43:52]:
The very bottom one says, has the number 413.

Jonathan Bennett [01:43:56]:
413. That's pretty good.

Ken McDonald [01:43:58]:
But that very last one is the catalog that I created in Calibre. called My Books, which is included in that.

Jonathan Bennett [01:44:07]:
It's a 412.

Ken McDonald [01:44:10]:
Actually, it doesn't show a 412, which means I had put one in and deleted it for whatever reason.

Jonathan Bennett [01:44:17]:
So just around 400. Yes,

Ken McDonald [01:44:21]:
that's just the paper.

Jonathan Bennett [01:44:24]:
Yeah, that's cool. I am tempted to do that. I've started at least once putting together a database of books. And, uh,

Ken McDonald [01:44:32]:
now one other one I did want to quickly cover using with this.

Jonathan Bennett [01:44:37]:
Yep.

Ken McDonald [01:44:37]:
That I like because I can set it up in a cron job is the calibre-check_library. I'm going to do help real quick. I'm not going to demonstrate it even though I did do it earlier just to make sure it would go quickly, but it would allow you to perform some checks on the file system representing a library. And it would be able to report invalid titles, extra titles, invalid authors, missing formats. If you've got Calibre on your system, open a terminal, do calibre check_library --help, and you can see everything you can do with it.

Jonathan Bennett [01:45:19]:
It's— but it's a great way— FSCK for your books.

Ken McDonald [01:45:24]:
Yeah, because One of the options that it does is vacuuming out the database.

Jonathan Bennett [01:45:32]:
Makes sense. Every database gets cobwebs, needs vacuuming out from time to time.

Ken McDonald [01:45:40]:
Every computer gets cobwebs, needs vacuuming out from time to time. Yeah,

Jonathan Bennett [01:45:43]:
that too.

Rob Campbell [01:45:44]:
All right.

Jonathan Bennett [01:45:45]:
All right, Rob, what about our networks?

Rob Campbell [01:45:48]:
What about our networks? So I have the command network Doctor, and apparently there's some more things I haven't figured out yet because when I looked, when I first discovered this one, it looked like a Tui app that you could actually do more, and I haven't figured that out yet. I don't know what I'm missing. So if someone could tell me, I've even asked the AM like, why am I not seeing the Tui? How do I get the Tui? I can't find it anywhere, but it still has a useful. things. So it's really kind of an all-in-one tool to like do nslookup, pings, all that kind of all in one shot. So if you just kind of do netdoc, um, netdoc by itself, it's going to, you know, do the checks and give you some details. It's like no obvious problem found. Okay, quit that.

Rob Campbell [01:46:39]:
Uh, if I want to do a check between a website, I can do an IP. Between me and a location, I'll do twit.tv, netdoc space twit.tv. And it quick ran through it. I guess you could barely even see it, but it had a list of things that it did. And it tells you things about it.

Jeff Massie [01:46:59]:
You might want to zoom the text a little.

Rob Campbell [01:47:02]:
I might.

Jeff Massie [01:47:09]:
Just trying to help you out here, buddy.

Rob Campbell [01:47:12]:
Don't worry about the text.

Jonathan Bennett [01:47:15]:
Nothing's going to happen. It's a command line tool. The text is unimportant.

Rob Campbell [01:47:23]:
All right. If you really think so, hold on one second.

Jonathan Bennett [01:47:28]:
We'll do it live.

Rob Campbell [01:47:30]:
I'll do it live. Let me just do a little bit of adjustment here. All right, maybe that's better, maybe it's not, I don't know. I'm gonna go with it. If I do a netdoc --watch to twit.tv, this kind of maintains a watch on it. It's a little bigger, it's still kind of small, but, and it's gonna do the thing, it's gonna just keep repeating through and watching to see if there's any problems it finds. And The other one is, I have an example here. If you do something that's wrong, because all those worked in the past, like, oh, I don't know, it looks fine, whatever.

Rob Campbell [01:48:14]:
So I just kind of add a whole bunch of Ts for twit.tv, netdoc twit.tv, and it tells me DNS twit.tv fail, cannot resolve. TCP fix, check the hostname. So it says a whole bunch of things that are wrong because it's not a real thing. So anyway, NetDoc is just kind of an all-in-one tool to check.

Jonathan Bennett [01:48:44]:
Document your nets?

Rob Campbell [01:48:46]:
Yeah, kind of find out where there's a problem if you're having something somewhere. And I guess you can go to details and kind of look around, but.

Jonathan Bennett [01:48:53]:
Yeah, I saw the TUI as well on their website when I looked it up and I was, I'm, I'm curious why you're not getting it. Maybe you're running that in a VM? Yes. It might have something to do with it.

Rob Campbell [01:49:04]:
I don't know. Yeah, because like the TUI, it kind of shows you you can go through and individually do things and do even more work. So I don't know what that is about, but apparently there's even— as great as this app is right now to just do a quick check on things, Apparently it's even better. And I don't know. Yeah. I actually put a question on the, on the, uh, in the, the issues track on GitHub for just questions like, why can't I get the TUI?

Jonathan Bennett [01:49:39]:
All right, Jeff, what you got?

Jeff Massie [01:49:42]:
Wipefs. So that's this week's, this week's command line tip. And it's one that I just used for when I did my, uh, RAID conversion. So when I moved my home server storage from MDADM RAID 1 over to a ZFS mirror, you know, when I had— I pulled each drive out of the old array and would hand it to ZFS. The problem is that a drive that's been in an MDADM array remembers it. There's a small marker written on the disk called a signature that says, basically says, hey, I'm a Linux RAID member. Partition tables, files, file systems, you know, leave their own signature too. So if you leave those behind, you know, i.e., don't take care of them, some tools can get confused.

Jeff Massie [01:50:27]:
And it's possible that ZFS may refuse to create a pool on the disk without forcing it. Or worse, something, you know, you reboot and it says, hey, this is an old RAID marker here, I need to put this back together and help you out. And it's like, uh-oh. Nope, I don't want that. So that's where wipefs comes in. It comes from the util-linux group of commands, so it's— should everybody basically should have this one. And just to play around with it, the first thing to know is that wipefs with no options is completely safe. So if you just type sudo space wipefs, W-I-P-E-F-S, And the device name like /dev/sdc, it's only going to list what it finds.

Jeff Massie [01:51:17]:
You know, for each signature, it shows the offset of the disk, where the markers live, the type, um, like GPT or ext4 or Linux RAID member, and it'll show you the UUID and the label if there is one, and it won't change anything. So It's, it's safe to play with that way. Um, so before you even plan to wipe a drive, run it just like that to see what's there. Are you sure you have the right drive? You know, um, you will need sudo just to even look because it reads the raw device. So even, even in the read-only, you got to be root. Uh, okay, you're sure of it. You add a -a for all, and that erases every, every signature wipefs can see. Now, when I used it on my drive for the first one, I did a sudo space wipefs space -a space, you know, /dev/sdc1, which was a partition, space /dev/sdc.

Jeff Massie [01:52:28]:
Which is the drive. Now notice the order there. The partition goes first, then the whole disk, and that's not by accident. It's— and it's actually the example if you look at it in the man page. Wiping the disk only removes the partition table. It doesn't touch whatever's inside the partitions. So if you just swipe— wipe the drive, The RAID marker inside the partition, you know, there's still one sitting there on the disk. It's just hidden until something finds it.

Jeff Massie [01:53:02]:
Probably not happily, or it won't end happily for yourself because it's going to then try to rebuild or whatever. So you do the partitions first, then you do the disk, and you've cleaned out everything. And WipeFS is smart enough to wait until at the very end to tell the kernel the partition table has changed. So listing them all in one command works. You don't have to do one and then wait and then do the next. Just string them all together and you're okay. Now, just so everybody's aware, and sometimes this gets people, but wipefs does not erase your data. It only erases those small markers.

Jeff Massie [01:53:46]:
You know, they're just a few bytes. They're not huge. things that it's getting rid of. Your files are still physically on the disk, it's just that normal tools can't recognize them. But that's why it's fast even on a multi-terabyte drive. But it means it's not a secure erase, and if you're recycling a drive or you need something secure, you got to get a different tool. This is, this is not it. And because it only touches a few bytes Mistakes can sometimes be undone.

Jeff Massie [01:54:20]:
You can add a --backup, and wipefs saves every signature it erases to a small file in your home directory named after the device and the offset. The man page shows you how to write it back with dd if you wipe the wrong thing. There's also a -n, which is the— basically, it's the dry run. And it shows— it basically tells you what you're gonna— what it's gonna do. A couple other things to note. Wipefs will refuse to erase a file system that's mounted unless you add a -f for force. You shouldn't be doing that. Unmount your drive before you really start tearing into something.

Jeff Massie [01:55:04]:
And just so you know, it's possible that drive letters like SDC can change between reboots. So before you wipe anything, check the drive's permanent ID to make sure that, you know, my SDC is really the drive I think it is. Especially on systems with a lot of identical drives, it's just cheap insurance. So double-check before you wipe anything. Take a look at the link in the show notes for many more switches and things that describe even more functions of WipeFS.

Jonathan Bennett [01:55:41]:
Yeah. I like the backup. The, you can probably fix this.

Ken McDonald [01:55:44]:
Is it wipe or wife FS?

Jeff Massie [01:55:48]:
Wipe. W-I-P-E-F-S.

Jonathan Bennett [01:55:52]:
Yes, indeed.

Jeff Massie [01:55:53]:
Like I wipe it clean.

Jonathan Bennett [01:55:55]:
Yes, wipe it with a cloth.

Rob Campbell [01:55:56]:
I wipe it clean.

Jonathan Bennett [01:55:59]:
All right. I do have one as well. This one is tied to our security story, and it's actually a couple of tools to check that. One is, it's emudra.wehaveaproblem.com, and there is a test here that runs right in the browser, and it will check your browser. Basically, it'll tell it to try each of these bogus websites. And I've got a couple that are timing out, but last time I did this, they all gave me the did not connect, which is good. That means that my browser is not connecting. It's not trusting one of these bogus certs.

Jonathan Bennett [01:56:39]:
However, if I go over here to curl and check these out, this one, it says SSL certificate problem. That's good. But this particular site, your client successfully connected. That's bad. So I need to grab updates on this laptop to get the latest certificate authority files where I'm sure this CA has been blacklisted there. But yeah, it's a really quick way to test whether the machine you're on is up to date and properly rejecting those certs and that CA. So really, really quick tip. Yeah.

Jonathan Bennett [01:57:28]:
And this was, this is something else that people have pointed out with this. Like your browser may get an automatic update right away, but what about the rest of your system? What about curl and wget and all of those things? So

Ken McDonald [01:57:39]:
what about some of those long-term supported systems that get updates once a month?

Jonathan Bennett [01:57:48]:
What about the firmware in your router that tries to reach out and get updates over HTTPS? Yeah, there's some, there's some really prickly problems with this. It's not good. All right. That is it though for the show notes. That is it for the tips. And yeah, we're going to let the guys cover whatever they want to close us out, get the last word in. We'll let Rob go first on that.

Rob Campbell [01:58:14]:
All right. Here we go with my usual closeout for the week. And it's a... If you want to have more of me, I've been posting a lot on the social medias. So to find me online, first, the easiest way at least, first go to my website, robertpcampbell.com, right here at the top of the page. Um, and from there, once you get there, you can, uh, there's a link to my LinkedIn, uh, Twitter, which I don't really use. Blue Sky, Mastodon, and a place to donate a cup of coffee to me. And if you want to know more to me, check out my book, rated 4.6 stars on Amazon.

Rob Campbell [01:59:01]:
It would be better except for Ken gave it a 4 only. 4.6 stars, I mean. Some really nice reviews.

Jeff Massie [01:59:09]:
I gave it a thoughtful review. Read the comments.

Rob Campbell [01:59:12]:
Yes. Very, very, very nice reviews on there. So nice that I've had somebody contact me and say they wanted to get my book out there more. Probably a scam. I don't know what it is, but—

Ken McDonald [01:59:26]:
Are they going to make a movie out of it?

Jonathan Bennett [01:59:28]:
No,

Rob Campbell [01:59:29]:
they're like, we can get you in some book clubs and stuff like that.

Jonathan Bennett [01:59:33]:
I always assume that those are a scam.

Rob Campbell [01:59:36]:
It was a really nice written email. I was like, yeah, they really nailed what the book was about and the things in it. It's almost like they read the thing. I don't know, came across your book and, um, it's so nice how this thing and you're this and it shows different stuff with your life and this and this stuff, blah blah blah. Anyway, if you want to check out my book too, you can click on the memoir link at the top or scroll down on my page where it's, uh, where there is a link. to buy on Amazon, or just go to Amazon. And if you type in Trailer Park 2 Technology Leader, usually it autocompletes once you get the 2 or start typing tech somewhere around there, and you can find it. It is available on softcover, hardcover, or on a Kindle.

Rob Campbell [02:00:28]:
I really— well, if I say this now, you're gonna be like, well, I'll just wait for that and not buy it now. You want this You want that too. I'm thinking about— I'm only thinking about, so don't wait for it. I'm only— but I'm thinking about doing an Audible version of this too for those who can't do it or be like Jonathan. He bought the hardcover. He has it. From what I know of him, I don't think he's read it. He's probably waiting for the Audible.

Rob Campbell [02:00:54]:
So I get it. You can do both just like that.

Jonathan Bennett [02:00:58]:
You got my number, Rob. I've got it. I've read a little bit of it, but Yeah, I've not gotten to much of it. All right, Jeff, what you got for us?

Jeff Massie [02:01:08]:
Uh, nothing, nothing too crazy. So I'm gonna have Poetry Corner. This one, this one's a little different. It, it is a haiku though. If you are British, this is a haiku for you. Aluminium. Have a great week, everybody.

Jonathan Bennett [02:01:28]:
That's, that's great. All right, again,

Ken McDonald [02:01:32]:
Well, I came across, uh, an effect that Plasma reimplemented back in 2024, and I did a recording of it and got a link to it in the show notes, along with a video I made demonstrating Amache's tiling window system that if you want to, you can watch it. I thought about using it in-house. I can still do that now, Jonathan, if you want me to.

Jonathan Bennett [02:02:02]:
I think we'll leave it there and let folks go and check it out in the show notes. We've been going for over 2 hours. It's time to go.

Rob Campbell [02:02:11]:
Yes.

Ken McDonald [02:02:11]:
Good news. They're less than 2 minutes each.

Jonathan Bennett [02:02:14]:
There you go. All right. Very cool. I appreciate you guys being here. Thanks for coming along and the fun we've had today. If you want to find More of me, there is, of course, Hackaday. And we've had a string of bad luck with trying to get FOSS Weekly recorded. We've had a couple of, one was a guest that got sick and then the very next week, miscommunication and didn't end up recording it.

Jonathan Bennett [02:02:36]:
Hopefully this week, we've got something really fun this week coming and looking forward to that. Hopefully we can actually make the interview happen. But yeah, other than that, just want to say thank you. Appreciate everybody that's here with you watch or you listen, you get us live or on the download, and we'll be back next week on The Untitled Linux Show. We'll see you then.