625: They're Doing it Wrong!

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:11):
Hello, friends, and welcome back to your weekly Linux talk show. My name is Chris.
My name is Wes.
And my name is Brent.
Hello, gentlemen. Well, coming up on the show today, we're going to chew on a bold proposal.
Should distributions stop packaging most of our desktop apps?
And why Flathub might be the key to Linux's future, but perhaps its biggest risk as well.

(00:34):
Then after that, we're going to dig into Linux 6.16, which is landing with killer
file system upgrades, some retro Mac fixes, you won't believe, and more.
They're going to round the show out with some great boosts, some shout-outs,
some picks, and a lot more.
So before we go any further, let's say time-appropriate greetings to our virtual
lug. Hello, Mumble Room.

(00:57):
Hello, hello, MiniMac. We don't hear you, but I see you. Hello there,
hello there. We have a small on-air, and then we have a delicate,
soft, quiet listening with a few folks up there as well. It's a quiet Sunday in the Mumble room.
Yeah, but we've got producer Jeff producing from the Mumble.
PJ, coming in from the mum. Pretty nice.
And a big good morning to our friends at Defined Networking.

(01:19):
Go to Defined.net slash unplugged and go check out Managed Nebula from Defined Networking.
It's a decentralized VPN built on the open source Nebula platform that we love.
We love it because it's optimized for speed, simplicity. It has industry-leading
security, and you can completely self-host it if you like, or you can take advantage
of their managed Nebula project.

(01:40):
If you go to define.net slash unplugged, you can get it for free on 100 devices.
Unlike traditional VPNs, Nebula's decentralized design keeps your network resilient,
whether you're managing a home lab or a global enterprise.
They built it to scale in 2017 for Slack's global infrastructure,
so it had to work from the beginning to connect their vast infrastructure together.

(02:02):
Nothing else matches Nebula's level of resilience, speed, and scalability.
Try it out for yourself first. Start playing around with it.
Host a lighthouse and see how that works. They utilize this really cool system for lighthouses.
It's essentially a node that contains the routing information for your mesh
network, and it's really easy to host yourself.
The queries from your network hit these lighthouses, and then you can have multiple

(02:24):
lighthouses as well, so you can have uninterrupted operation.
So if one goes out, you have another one going, You're still going to be able
to connect to all your devices.
Of course, they have the managed offering as well, which handles a lot of this for you.
Nothing offers Nebula's level of resilience, speed, and scalability.
Go see why we're switching to Nebula and try it for free for 100 hosts.
No credit card required at defined.net slash unplugged.

(02:50):
And thank you to Define for sponsoring the Unplugged program,
Define.net slash Unplugged.
So this week we're going to talk about something that's kind of far out there,
and it's a hot topic right now because the Fedora project is considering a proposal
of how they're going to handle flat packs from FlatHub and from Fedora directly

(03:12):
from Fedora 43 and beyond.
And in the immutable desktop world or the atomic desktop world,
this is very important because this is essentially how you distribute modern
desktop software is through flat packs and for those of you not aware there's
flat hub which is the public flat pack database and then there's fedora flat packs which.

(03:35):
Is a flat pack repository but maintained by fedora.
And currently there is a proposal titled filter fedora flat packs for atomic
desktops And the summary says, quote,
With this change, we want to make the availability of Fedora flatpacks an explicit
decision for Fedora Atomic Desktops.
Fedora contributors may package any application as a Fedora flatpack,

(03:58):
but those flatpacks will not be made available immediately to Atomic Desktop users.
Users that want to have access to all flatpacks from Fedora can remove the filter.
there's a debate going on right now if fedora should even be
in the business of packaging flat
packs there's very little maintenance that they see very little community engagement

(04:20):
they haven't even held some of their main meetings this is not a criticism because
they're a busy project there's a lot to do but there's very little documentation
how to build maintain or update fedora flat packs there's no procedure to remove
deprecated fedora flat packs,
And then building a Fedora Flatpak seems to be a different process than,
say, an upstream Flatpak that builds for Flathub. So there are different processes.

(04:41):
Oh, yeah, right. I mean, they're totally disconnected. You're using all your
regular RPM system and Fedora tooling to go build this version versus,
well, whatever Upstream's doing.
It does have some advantages. Fedora Flatpaks pay a lot more attention to free
software and make sure that they're building directly from the source.
They're built using the same source RPMs that the Fedora distribution packages are.

(05:03):
So, you know, if they're doing some patching there, something like that,
that gets included. So there are some benefits, but they're not getting particularly well maintained.
And they seem to be creating user frustration. There was an LWN article in this
realm, and some of the user comments really stuck with me.
One user wrote, I had to step through a user who was having a bizarre issue
with a package that I couldn't reproduce on the same Silverblue system.

(05:27):
After several comments and hours of debugging, I realized the user had mistakenly
installed the Fedora Flatpak version of a package that I had installed from
Flathub. The Flathub package worked fine.
They were frustrated, as I was, when they realized this, and they immediately
removed the Fedora Flatpak repository and reinstalled everything via Flathub.

(05:48):
And there's a couple examples of this. It creates frustration for the user. There's confusion.
There's definitely confusion, yeah.
We always see this a little bit when you do have separation between upstream
and who's doing the actual shipping to the end user.
But that's maybe where some of the like how much focus and bandwidth is available

(06:09):
for that as well because it's even worse if you don't have anyone there to actually
go figure out like, oh, is this a downstream issue or no, you should talk upstream.
The advantage that the Fedora flatpacks offer the Fedora project is they don't
have to worry about some of the legalese problems that might be in FlatHub packages
and things like that. It really always comes down to that.
Well, they have full control. The providence, the security can all be,

(06:33):
at least to more of an extent, vouched for or changed.
And when that's often sometimes, you know, for especially for like a common,
maybe in the future desktop apps, maybe there's concerns there around two of
what, what, what software do you show and sort of represents you and what control
do you have over it? I can understand that.
It's just if they're universal packages until they're not, and people are packaging

(06:56):
them differently and they have different repositories that overlay on top of
things and there's filters on top of this and that.
And so when you go to the website, you get different results than you do in the software center.
It's not a good user experience. And I think Flathub is becoming a bit of a
dominant player in this space.
I was just looking at their stats, which they make publicly available.
They're approaching, they're just on the cusp of 3.1 billion downloads.

(07:21):
They have 3,065 Linux desktop apps, of which 1,657 have been verified.
I don't know about any public stats for snaps and app images,
but that's a powerful presence.
And what I see here is a lot of duplication of effort.
And before I go any further, I want to just put this thought experiment out

(07:43):
to the audience and have you chew on this and tell us what you think.
Boost or go to the contact page.
As we go through this, think about this deeply.
If you're distributing switch to Flathub-only desktop apps, I'm not talking
about the core utils, talking things like Firefox, your text editor.
the things in the desktop email client if that all came from FlatHub as flatpacks

(08:08):
or snaps or app images what specific tool or workflow would break for you in that scenario.
What changes? Because let's consider a core problem that we hear over and over again.
Free software developers don't have enough time, don't have enough resources,
don't have enough money.

(08:29):
And then when you consider that Flathub is at almost 3.1 billion downloads,
or that things like the AUR, which are user-contributed packages,
can be the make-or-break success for certain distributions, I guess it begs a big question.
Should Linux distributions be packaging nearly everything that they are?
They're still packaging the entire world.

(08:51):
Is it time to revisit this classic role, reduce redundant efforts?
Some of that could be redirected to improving Flathub and Flatpacks.
And these distributions don't have to hold these massive repositories.
Like, this just put ClearLinux out of business.
The cost of ClearLinux repositories was one of the key contributors to having
to shut down ClearLinux.

(09:12):
So I want this to be out there as a thought experiment as we go through this,
because there was a really good blog post.
Brent, I think you got to take a crack at this last name here.
It's Michael Contazaro.
Oh, I think you nailed it. That's what I was going to go for.
No, really?
Contazaro. Why not?

(09:33):
Well, he had a really thought provoking blog post, and the title was Fedora
must carefully embrace FlatHub. And he argues that Fedora must strategically
shift towards Flathub to fully realize the benefits of an image-based,
Flatpak-centric desktop future.
And before you go, well, why is that the assumption?
Fedora has stated Silverblue and Kino Knight are the future of Fedora in their

(09:53):
2028 strategy update outlook.
They write in there, Silverblue and Kino Knight are ready to be our desktop additions with Boot C.
We think image-based operating systems are the future. Let's commit.
So they're the ones saying it, not me,
But I happen to agree. After using Bluefin and Aurora and now Soltris OS,

(10:17):
these things are robust. They're hard to break. You can throw a lot at them.
This is clearly what end users and servers should be running.
And realistically, probably the future of RHEL, right? If this is where Fedora goes.
Well, and I mean, we've already just seen, right? RHEL's pushing on image mode.
Yeah, and they're very pumped about Bootsy. Now, it doesn't mean that RPM-based

(10:38):
distributions are going away. I think it's like a flippening happens.
The image based ones become the default and the RPM based ones kind of become
like the alt that's available.
Like how silver blue is right now, it flips and workstation RPM based is the
et cetera download and silver blue and Kino night are the primary.

(11:00):
That becomes the first thing you get to the one that's viewed as primary. It's a big change.
Yeah. We're not ready yet, but I think we could get there pretty quick,
actually. That's one of the things I've realized recently.
And I think the author is right in this post that Fedora Workstation,
as it is today, for a regular user, it's breakable.

(11:21):
Like if he gave it to a YouTuber, they'd break it.
They'd try to install Steam and uninstall their desktop. Something like that.
I really think it's better for users to switch to these image-based systems.
And I feel really strongly about this. And in that world, you don't package all these RPMs.

(11:42):
That's an old practice. Right.
I mean, just for a variety of reasons, but especially if all you're really using
the RPMs for is to build the base images, then you only really need the set of things in there.
You need to create a robust base Linux.
And I'm just trying to think, where else in history have we seen 3.1 billion

(12:03):
downloads from what is essentially a Linux app store?
FlatHub is the biggest success in this desktop app store-ish space that we've ever seen.
Well, I remember when it kind of switched, you know, we clued in that like,
oh, we could go find some good picks and other cool software from just browsing
what's new and popular on FlatHub.
And I also find the Fedora Flatpak experience bogus. And then you,

(12:25):
so you have, you know, everybody out there is packaging their own version of
Firefox. It's kind of crazy.
You know, if you think about it, if you're running Nix OS on one machine,
Ubuntu on one machine, and Silverblue, and then Fedora on another machine,
each one of those is technically running a slightly different version of Firefox

(12:46):
that has been packaged by somebody upstream.
That's bonkers if you think about it. And it's not how it works on macOS or Windows at all.
And it means all this work to repackage Firefox and TextEdit and Console across
all these distributions all the time.
It seems to me like we're stuck in this way of doing it because in a previous

(13:10):
life, a generation or two ago, how a distribution packaged software,
the way they packaged it, the way they broke it up, their default configurations,
that was one of the leading factors in how you picked a Linux distribution.
And it really mattered when software was being shipped on CDs.
Well, and just even the act of distributing it was a service,
right? Like it was either that or you're going to go build Firefox for yourself.

(13:32):
Because they're all like TARS on FTP servers.
Right.
And it truly, like the fact that you could get it in binary form and have these
repositories you could pull from was a genuine, unique value offering.
And maybe there was more too, you know, in terms of tying things together.
Maybe there was more patching or tweaks or initial configuration needed to make
it kind of be able to play nice with the rest of the desktop.

(13:54):
That's gotten a lot better too.
I think that culture is just so ingrained into us in Linux that we don't even
realize we're still building it that way, even though we don't need to anymore.
And I just think about the duplication of resources there.
and you know the status quo right now say on fedora workstation i'm
picking on fedora because i think fedora is actually becoming something kind
of special right now when you

(14:17):
set up fedora and you get it loaded it then comes up with a screen that says
would you like to set up fedora flat packs fedora rpms and the open h264 rpm
and there's this like if you and then if you don't like turn on the flat hub
then the next time you open up gnome software it'll then suggest it again,
and it'll also add a couple other RPM repositories.

(14:38):
And it's just such a bonkers out-of-the-box setup already.
And I don't think sticking with Fedora flatpacks in any way is going to make this better.
And maybe, you know, you package a few base must-have things and then all the
other stuff you pull from FlatHub.
Yeah, that's kind of this post-author's opinion, right?

(15:00):
Like maybe just have the stuff that you install by default,
you know that kind of represents you that's your public face that
you really need to be rock solid and maybe have a little more control over use
fedora flat flat pack for that but then yeah anything else that you want go
to flat hub although i think crucially right there's also a recognition here
that it's gonna if you want flat hub to improve enough to be suitable for this

(15:23):
purpose there's a lot of work to do and probably the fedora community should try and help,
And I mean, of course, the wider Linux community, I think we would argue too.
If this idea of, you know, a distribution having its own private flat hub offering
for its default applications, if every distribution then has its own,

(15:46):
don't we end up with the same like PPA problem that we've had in the Ubuntu
ecosystem where there's some installed by default?
And if you want what everybody else is using, then you end up with conflicts
or something like that. I don't know if that's really the future I want.
I agree with you there. And I think this is where distributions differentiate.
And they have different approaches on how they package the base system that you get.

(16:10):
Because you need to still have the core utils installed. You need to get the
Linux packages installed and all the stuff around that.
You've got to get all that set up. And you've got to have some base apps that
work right out of the box.
So I appreciate that problem.
I mean, yeah, there's the question of, like, where does your graphical terminal
come from? Do you have to ship that from the FlatHub version?

(16:31):
Well, Bluefin does.
It's true.
Aurora does, and it works great.
But that doesn't mean it's the only way that can work well.
I think it could be a Blueprint, though. I really do think they're onto something
here. And it also makes my system a lot more portable.
Like, I can rebase to different systems, and all my user apps stay there.
I think the core problem here is, is FlatHub ready for a responsibility like

(16:52):
this? That, I think, is a wider question, right?
Well, I just mean like if you, you may also want more gated controls and what
counts as a user app, right?
Why would you want more gated controls? Just because it's legal?
Well, no, no. What if upstream breaks your terminal that you just,
just after you did your most recent release?

(17:14):
We're going, you know, Linux Unplugged goes and does a review and the terminal app doesn't work.
But isn't that a problem?
I'm just arguing that there's, that's a debate of like, do you have some level
of control? because like that's a bit different than you know.
D packages you're presuming the packager has tested in a way that would catch
that before it goes downstream like the rpm version yeah i don't know if that's true for.

(17:35):
Like some of the stuff on the yeah.
I mean some of it would be but you know and then if.
You do any sort of release qa that's the kind of thing.
You would.
Launch the terminal.
That i so i think right so there there's a few things here that would have to
work you'd have to have some pretty robust testing,
Flat Hub would have to be probably a little more robust, right?

(17:55):
There's some good criticisms that we can get into if people are interested.
But if we're going to get into the Flat Hub criticisms, I'd like to bring a
Flat Hub person on board just to respond to them before we just blast a bunch of criticisms at them.
But some of them are pretty obvious. There's a thousand apps that run on end-of-life
runtimes, and the reasons for that are various.
Some of them are because they're not being maintained, but some of it are because

(18:17):
of compatibility reasons, like with NVIDIA drivers or whatnot.
And there are some apps that try to subvert the sandboxing or have overly broad
permissions, probably needs tightening up.
And we're still, I mean, it's been a long voyage, right, of working out all
the right holes and protocols and interfaces to, like, some apps couldn't be
sandboxed and still properly function.

(18:38):
Right.
So we're still kind of fixing some of those, too.
And I still think it's better than no sandboxing.
Yeah, for sure. Especially when you're still in proprietary apps.
The idea that the blog author floats it I think works kind of well, And I'm curious to know,
Brent, if there is maybe a middle ground here that you think would work is Fedora
clearly isn't capable or isn't comfortable just letting the whole hog flat hub

(18:59):
install anything you want that might be, you know, technically not legal.
That that bothers them. You know, they've always had to walk the line with legal.
This post author suggests that.
Gnome's got like a probably safe rating that they're introducing for applications.
You could essentially label things as probably safe, and maybe that's all the filtering you need.

(19:21):
So instead of right now what they have is an allow list that is essentially
blank at the moment, but they can block and allow certain apps from Flathub
with a pre-filter list that they have.
They're not really utilizing at the moment. There was a discussion around that.
But what if that pre-filter was just simply, by default, just show the probably
safe apps, And then there's a checkbox or a drop-down menu that you uncheck

(19:43):
that shows all the other apps if you want it.
Couldn't that be a pretty simple middle ground where they don't have to actively filter?
It's something that needs to happen anyways on the FlatHub side,
and the GNOME project already has some efforts around that.
Maybe this is a middle ground? Or is filtering just always going to be a lost cause?
Well, I feel like some visibility into the trustworthiness of a Flatpak is always good.

(20:06):
Everybody wins with that. but you end up with
this workload of someone has to go through
them and trust them and revisit
that trust on a regular basis and who
is responsible for um applying
that trust right is it the gnome project is

(20:26):
it like a bigger conglomerate of
like a bunch of distributions coming together with representatives that
filter through these so some transparency
i think it's really good this is a problem many distributions would want to
solve and me and as an end user and i think you guys too having more visibility
into the trustworthiness of uh you know publicly available flat packs on flat

(20:52):
hub is a definite win it's one of my huge hesitations with using flat packs for everything,
So, yes, I think it's super important, but it's not necessarily obvious how
to accomplish that in a way that isn't a bit nebulous.
And a burden on one particular set of people or something too.
Yeah, I'd love thoughts if people have thoughts around that.

(21:14):
I think the other thing that I want to put out there as a thought experiment is what if, just what if,
somewhere between 20% to 50% of the distro packaging maintainership out there is redundant work?
What if 20% to 50% of the apps that they're packaging are getting packaged by

(21:37):
their counterparts at 200 other Linux distributions?
Doesn't that seem like a wild misuse of resources and time when we have solutions now in place?
And even if some of us wouldn't prefer them and we would use other tools like
the AUR or the Nix package repository or maybe, you know, Debian would probably
never go this way. So you'd still have all that.

(21:57):
But like, you know what I'm saying? Like, it just seems even if it's 20%,
it's something worth talking about.
And then here's the other realization I've had in the last few weeks since we were in Boston.
using Bluefin and Aurora and Soltris OS.
Bluefin doesn't have a package repository at all.
The Bluefin devs are not wasting their time repackaging Firefox and all the

(22:21):
other apps for the 200th time.
And what I've realized is Fedora is about to enter their Debian Ubuntu moment.
We've said it on the show before, back in the day, why aren't there more distributions
based on Fedora like there's distributions based on Debian.
There's so many Ubuntu and Debian-based distributions and Arch-based distributions,

(22:42):
it's really kind of wild. But you don't see it around Fedora and SUSE as much.
Well, now we are seeing it around Fedora.
And Bluefin comes with a bunch of sane apps that humans would want to use with
sane defaults, just like Ubuntu did back in the day.
And it lets you just get to work. Silver Blue feels like installing the Debian

(23:03):
desktop after Ubuntu came out.
And it's got this weird kind of aesthetic.
It's got a super limited selection of applications. When you search,
the repositories feel really limited.
And it's this weird mix of new and old where you go over to Ubuntu and they've
curated a theme and a default selection of applications and settings for the

(23:26):
whole system that are very reasonable.
And app availability plays a huge part here because everything that's in Flathub
is available to me. And they have Bazaar or Bazaar, however you say it now,
to search apps. It's super fast and lean and mean.
And so while I do have a few FlatHub concerns, like old runtimes,
you know, and the fact that they do seem very GTK4, LibWadia,

(23:49):
GNOME focused, and, you know, I'm a Plasma guy mostly.
Like those are concerns I have, a bunch of other ones. But the difference is
so stark. and I don't think Fedora doesn't have to worry about it right now
because the UBlue based stuff is what, around 25,000 users total,
difference there is really something if you try silver

(24:10):
blue and then try bluefin it's very much an ubuntu debian experience and i think
fedora has to consider switching the way they deliver software to stay relevant
and stay competitive in this new future especially if kino knight and silverblue
like they say are their stated future yeah.
I mean if so yeah i think the things we're talking about here exactly what you

(24:33):
need to do to prepare for it so that your your base is well-optimized to live
in a future that is based around Flatpak.
Let us know, boost in, or go to the contact page if an app package,
if app packaging stopped being a differentiator, so say they mostly just went
with FlatHub except for the core stuff.
At that point, what would make you say pick Fedora over Debian or vice versa?

(24:57):
Say in the year 2028, they started implementing this stuff next year.
By the year 2028, most of your desktop applications are delivered via Snap or
AppImage or Flatpak, then what makes you pick a distro?
Are you looking for a distro that doesn't do that? What differentiates the distros now at that point?
Especially when you consider SystemD and the kernels kind of the same across all of them.

(25:21):
I'd love to know your thoughts, especially those of you that have been using
desktop Linux for a while.
Please boost in, support the show, or go to linuxunplugged.com slash contact.
1Password.com slash unplugged. That's the number 1Password.com slash unplugged, all lowercase.
Go take the first step to better security for your team by securing credentials

(25:44):
and protecting every application, even the unmanaged shadow IT.
If you're in security or if you're an IT professional, you know the mountain
of assets these days to protect.
From devices to identities and applications, It's just more and more all the time.
And with that, often comes a mountain of security risk.
But you can actually conquer that mountain of security risk with 1Password Extended Access Management.

(26:10):
You see, over half of IT pros say that securing SaaS apps, that's their biggest challenge right now.
And I can understand why, because it's just growing at such a rapid pace.
You could kind of call it sprawl.
And from that, you get your shadow IT, the stuff that's not even on your radar.
It's not hard to see why. Users are just trying to get stuff done.
That's where Trellica comes in by 1Password. It will discover and secure access

(26:34):
to all your apps, managed or not.
Trellica by 1Password inventories every app in use at your company.
And then it has pre-populated app profiles that can assess the SaaS risks,
letting you manage access.
You can also optimize your spend and get a handle on that, even quantify it.
And enforce security best practices across every app your employees use,

(26:55):
even the ones you didn't know about initially.
Now, 1Password came along, and they just changed the password game for managing
your identities online.
And they've done it in a way that's very approachable by end users and manageable by IT.
Well, that only a million times more, right?
That award-winning strategy now applied at the application level.

(27:18):
You have to really go check it out. So go to 1Password.com slash unplugged.
That's where you have to take the first steps because there's more to secure than just passwords.
Managed and unmanaged SaaS applications, for instance, are a big deal.
Trelica by 1Password will secure your apps without leaving your employees behind.
There are a lot more options now, and Trelica by 1Password can help you take

(27:40):
control of your SaaS environment.
Check compliance off your list and reduce unnecessary costs by analyzing app
usage and using data to identify and eliminate unused licenses and redundant apps.
It really, truly does make things better for both you and the end users.
Check it out and support the show by going to 1password.com slash unplugged.

(28:10):
Okay, well, you can tell a kernel release is a good one.
When checking my notes here, I'm posting at five minutes after midnight about,
oh my gosh, there's just so much great stuff in 6.16.
Is that Larble?
No, that was me.
Oh, that was you? I could see being either one.
That was after me reading like most of the changes in the kernel and having

(28:31):
a hard time picking which ones I wanted to even attempt to slot into the show.
I agree. This is one of those where we could go on for two episodes about all
the new stuff because NVIDIA Blackwell saw a bunch of improvements.
The new Intel stuff, RISC GPU drivers saw a lot of work. Even Ethernet drivers.

(28:51):
Virtualization improvements. Just a whole raft of stuff and,
of course, some new Rust stuff as well.
Like new Rust kernel driver bindings and things like that.
Yeah, the Rust effort marches on. More subsystems get in interfaces.
I don't know what it is about the three of us, though, but something about the
file system news just really seems to be the most exciting to us.

(29:12):
I think it's, I blame Brent because he's always talking about building a file
server, so it's your fault, Brent.
Well, I know how to get you going.
He's perennially doing a backup, so.
It's, I think also, it's amazing the amount of improvements and work that are
landing across the various file systems still.
Everyone wins with 6.16.
Yeah, really. ButterFS continues to get faster in 6.16. The biggest highlight

(29:35):
this cycle is performance for ButterFS.
There's been buffer conversion work that helps metadata-heavy operations,
so you get throughput and runtime improvements.
Yeah, plus 50% throughput, minus 33% runtime. That sounds great.
On top of that, cleanups and efficiency tweaks, they say, So some just nice
overall gains from that. It's good to see that too.

(29:57):
Leaner and meaner ButterFS in Linux 6.16, even when dealing with lots of metadata.
ButterFS has been going through a six plus year cycle of improvements in every Linux release.
I mean, think about how that stacks up over time. So if you haven't tried ButterFS
yet, your perception might be a little out of date because I'm talking literally

(30:21):
every kernel release for the last six years, has had really nice butter FS improvements.
What hasn't seen a lot of improvement for a while, because it's just pretty
stable, but saw some big changes this release was...
That's right, Extended 4.
It's been a minute, but 6.16 introduces notable performance enhancements for
Extended 4, primarily focused on three areas, fast commit improvements,

(30:46):
multi-FS block atomic write support, and big ALEC file systems,
and most significantly large folio support for regular files.
Yeah, as Ted So wrote, this last result there, the large folio support,
it can show really stupendous performance for the right workloads.
Stupendous, huh?
Yeah, stupendous.
What is stupendous? Can you measure stupendous?

(31:08):
For example, see, he links here, we'll put that in the notes,
where the kernel test robot reported over 37% improvements on a large sequential I.O.
workload. nearly 40 percent intel's kernel test robots so props to them.
So we're going back to Extended 4, boys?
I guess so.

(31:29):
Scrap your head now, Brent. We're starting over. We're going to Extended 4, okay?
You can do both, right, I think? Can't we do both?
Yeah, we'll do LLVM on top of it. It'll be fine. LLVM on top of it.
Last but not least in our batch of updates, well, not last, but almost last,
is XFS. So second to last, but still.
XFS officially merged Atomic Write Support.

(31:52):
That's a big one.
Yeah. I believe it was developed by an engineer built upon groundwork done by
another engineer at Oracle.
So some Oracle time went into that.
It enables atomic writes primarily for single block operations due to limitations
in alignment and extent coverage.
A software-based fallback method is available and introduced for misaligned
or multi-extent writes, leveraging XFS's ref link copy on write mechanism.

(32:19):
I know, right? Even XFS is partially copy-on-write these days.
Hmm. Interestingly, tested it using MySQL databases and saw some nice performance improvements.
Additionally, experimental warnings for features like the PNFS and Scrub and
parent pointers features have been removed in 6.16, so they're no longer considered
experimental features.

(32:40):
Oh.
So XFS has Scrub now. It's got copy-on-write. It's got atomic write support.
Yeah, that makes sense, too, especially for databases. This kind of thing is
definitely nice. So XFS is an excellent database backing system.
You're talking a super high integrity file system now with reliable write consistency.
That's just great to see.
But do I want the atomic write support in XFS or do I really want to try out

(33:03):
the stupendous performance in the XT4? How many file systems am I going to have to make here?
Well, let me throw one last one at you. Our buddy, Kent Dover Street,
of course, got some BcacheFS improvements in 6.16.
And you know we love following this. Key highlights include performance improvements
with faster snapshot deletions, faster device removal, and coalesced accounting updates.

(33:24):
Yeah, basically it's kind of when you have metadata and accounting updates,
you batch them together before going to the transaction to commit to put a little
less pressure on the journal, which should just generally be good for performance.
Also in 6.16, there is significant progress on the self-healing and recovery
features of BcacheFS with background recovery passes triggered upon error detection

(33:45):
and a new recovery pass for the rebalance of Btree.
So it detects it in the background and initiates the repair.
Yeah, Kent's pretty much just been in bug fix and repair mode.
Error message enhancements, better error messages.
Yeah, just everything in, like, making it more robust and resilient and better
at telling you what's happening and what's wrong.
And more stats. I mean, there's stats encounters all over the place.

(34:07):
I don't know what this means for 6.17.
No, I don't think anyone does. I mean, Kent and Linus might, but we do not.
But also it was nice to see Linus merging a last-minute fix in 6.16 for a user-reported
bug, resolving an issue with encrypted file systems.
Yeah, this is all very nice to see because I think it just means if there is

(34:30):
going to be uncertainty around how you should be getting your BcacheFS,
6.16 will be a really great kernel to be on for a while until that shakes out.
This also marks the end of that little pause of BcacheFS being merged into the kernel. Is that right?
Well, it kind of marks this. Well, so it's marking the start.
This that drama unfolded throughout the beginning development of 616.

(34:53):
I got it.
And so what Linus kind of said is, all right, I'm going to accept these packages.
And then he said, we're going to part ways after that.
So I think what Linus and Ken are both kind of maybe offline or just have come
to some sort of a quasi agreement is they're trying, like Wes said,
get it into a state where if it gets removed from the Linux kernel for a while,

(35:16):
you're going to be OK on 6.16 until things get resolved.
And now the pause begins, but we don't know what that pause looks like.
we just have to wait and see I've made my feelings pretty clear about it I think
bcachefs is a critical Linux,
feature, probably more important than some of the other things that are getting
worked on right now in my personal opinion, but what the hell do I know but

(35:36):
we'll keep an eye on it and see where things go for 6.17, so it's nice to see
bcachefs get in a good state as good as can be at this stage,
before that happens.
Yeah, Ken's definitely been putting in a lot of hard work.
One of the distributions that's jumping on Linux 6.16 really early that I didn't expect is Ubuntu.
Kinda. You see this? This is interesting.

(35:59):
I guess for their concept ISO, which is for the Snapdragon X1 laptops.
Yeah, yeah. Yeah, well, 6.16 has patches and fixes for some of these Snapdragon
SOCs and the components in there.
and 616 adds device tree support for the
dell latitude 7455 the inspiran
7441 and the xbs 139345 now

(36:23):
you have a working fingerprint sensor um in there
and also i guess the zenbook a14 received some fixes so i
think ubuntu was just kind of motivated to ship this pretty fast um which is
great i guess also there was a fixed because uh grub was having trouble supporting
systems with more than 32 gigs of ram you run into these things in the arm land

(36:44):
because we haven't had a lot of powerful arm boxes until recently no.
We have not.
And then boom all of a sudden we're shipping them like real computers which
is nice to see uh so if you have one of these cutting edge arm devices and you're
running ubuntu you should probably be trying out the concept iso and.
Maybe you're telling us how it goes because i'm real curious.
Okay so i.
Know what you you have clearly have a favorite here.

(37:05):
There this one's really great um i don't know there's something about just getting
a little bit out of these old computers that I'm a big fan of.
And it's something that just sat for a long time, didn't see any love.
I think somebody discovered it in Linux 2.6.12. They discovered this issue in June of 2005.
It was a problem that was really only cropping up on vintage Macintosh 2s.

(37:31):
What?
Yep. It stemmed from an off-by-one error that caused the system to report a
Mac 2 as just a Macintosh unknown. Apple Macintosh unknown.
What a bummer of a bug, right?
You go to all the trouble of running Linux on your Mac 2, only to not really
get it to display correctly.
Yeah, and the system can't recognize there's no to do with it.
So a Linux hobbyist using a real Macintosh 2 discovered the bug when booting the Linux kernel.

(37:55):
The fix was merged alongside another minor Motorola 6800 update.
For us in the know, that's the M68K line, including a little conversion and
update for default configurations to the strscpy function.
but it's pretty cool so it fixed this old bug that's been around since 2005
this off by one error that caused the old Mac 2 to be labeled as Apple Macintosh

(38:18):
unknown now it gets properly labeled,
and there's just something I think really great about seeing,
Linux still getting work for these old retro systems it just worms from hurt yeah.
It's just kind of cute.
And while we're cutting you.
Know there's still a little bit of community spirit left in this corporate kernel of ours.
Yeah yeah Maybe some places don't, you know, they don't want 32-bit support

(38:41):
anymore, but it's nice to see if you could get it working. It kind of makes me want to do it.
I'd love to try to get an old Mac running Linux and then figure out a way to
get a web browser or running like a web server on it, making a server.
That's what I'd like to do. And then open it up to the audience. Come on.
You know, serves your personal web page.
So you did a little bit of digging there, Westpain, and you just found a whole

(39:04):
bunch of links that we can throw in the notes. Some of it just like good background
on the 6.16 development work cycle.
But you got some specific stories and a few findings. Anything you want to highlight before we?
Yeah, well, I kind of just, yeah, I did. We'll have some links for just if you
want to review this stuff. And then I kind of dug through, like LWN had a great list.
Phronix, of course. And then Colonel Newbies is out ahead on 6.16 too.

(39:25):
So I just have a little lightning round of neat stuff, I noticed.
To start, Core Dumps can be sent to a Unix socket instead of just being written
to a file or spawning a user mode helper.
and this is part of a bunch of work to try to make the core dump interface more
secure because i guess there's been a flow of core dump related cves and some
of the kernel developers are getting a little fed up with that so that's kind of interesting to see,

(39:48):
You were speaking about I.O. Euring.
In the bootleg pre-show.
Indeed. Well, I.O. Euring can now be used to create pipes. Yeah,
that's right. Super fast pipes.
Hey, I like fast pipes. I'm all about fast pipes.
There's now something that you can do called zoned loop block devices,
which emulate a generic block device using multiple files on an existing file

(40:08):
system to basically let you emulate zoned devices for testing,
for whatever you need to do with that. That's kind of neat.
Hmm. Well, that's something we might be able to play around with.
Yeah, it kind of sounds fun. And, which this does not always happen with new
kernel features like this, there's some pretty nice documentation that exists right out of the gate.

(40:29):
Yeah, I was just looking at that right now. Like, get up and running with it
kind of documentation and how to use it. That's fantastic.
Here's something you hopefully don't actually have to watch out for,
but, you know, like we say, don't break user space.
6.16 is actually removing a system call, the ancient use lib system call.
It's been deprecated for quite some time.
It's actually been removed, and as they say, hopefully without breaking any

(40:54):
user space applications. It's just you don't see that all the time, so watch out.
I did notice there's a bunch of audio hardware support that's coming along,
but including a six-channel professional DJ mixer.
Whoa, really?
The DJM V10 from Pioneer.
Oh, let's do it, boys. We'll get a mix

(41:14):
going. Brent could be on the drums Soundboard guy could be on the mixer.
That's right This.
Is actually a really nice looking.
It seems like a good rig Yeah it does I mean I don't DJ but No It looked real nice on the shelf Says.
It has elite sound quality.
Okay this part's getting me really excited Alright There's a new API That will
let virtual memory allocations Persist Across KExec handovers Oh,

(41:40):
Yeah Uh huh This could This could this could do a lot because now you kind of
have like an official way for the two kernels to like pass information.
Oh man.
I've been emulating this with my like persistent memory hack sort of thing for
file systems but you know having it be first class in the kernel could lead
to some really cool tools.
When we had the disposable system that was created through KExec that was one

(42:01):
of the things that we kind of had to work around.
Now we could just have that functionality built in. We're definitely going to
play with that in the future. Yeah.
I found something for Brent here. Crash dump kernels can now reuse existing
Lux keys, which means crash dumps can actually be made to encrypted file systems
now, which was not previously possible.

(42:21):
That is big.
Yeah. And you can imagine it probably took some careful work to get that all
aligned without breaking anything.
It's a feature I didn't realize I needed.
That's right.
You need to save your dumps, Brent.
Maybe you've used like the ffa notify command where
you can like watch for changes in files there's now
a feature where users with the cap sysadmin capability in

(42:42):
a user namespace but no special permissions in the root namespace can watch
file systems and mounts with fa notify so now you can have like containers in
a user namespace they just need cap sysadmin and then they're allowed to like
watch all their own files it's
just beefing up the things you can do with containers securely interesting.
Interesting way to go about getting that information to the container okay.

(43:05):
Usb audio devices now support audio offloading this lets for example audio from
a usb device continue to flow even when the rest of the system is sleeping what whoa yeah,
so you could like basically tell it where tell the network stuff i guess like
where in memory to read from the USB and have it just flow with the rest of

(43:28):
the system, not needing to do anything.
And then Greg KH pointed out that I think this takes the record for the most
number of patch series, 30 plus, over the longest period of time,
two plus years to get merged properly, at least for USB things.
I mean, that is so fun and neat and awesome. I want to play with that.

(43:49):
See, this is, I am getting excited and nerded out by Linux kernel features.
This is a good batch This is a good batch This.
One is going to be big for those AI workloads The contents of device memory
Can now be sent via TCP Allowing zero copy transmission From a GPU to the wire.
The contents of device memory Can now be sent via TCP That's crazy That is huge

(44:13):
for AI workloads They're going to build on top of that immediately Yeah.
I think it's been some people been asking for.
Allowing zero copy transmission From a GPU to the wire,
Whoa, there is, that's a big feature.
Something closer to home. I know we're all, you know, we love things like WireGuard and Nebula, but.

(44:34):
What'd you say?
WireGuard.
There you go.
If you are an OpenVPN user, that's going to get better with 6.16 because there's
now a virtual driver for offloading some operations that OpenVPN does to the kernel.
That will make a difference. So the, like, is it probably encrypt stuff that's
going to be offloaded to the kernel?
Yeah. Let's see. I do have.

(44:57):
I know. I should have looked myself. So they're creating a virtual driver for the data channel. I see.
I guess that makes sense why it would impact large transfers. That'll be nice.
Okay. You know we love eBPF here.
Yeah, buddy.
You can do traffic control, right? Like traffic control stuff for queuing and
making your network play nice, quality of life kind of stuff.

(45:20):
you can now implement custom traffic control queuing disciplines,
basically like the algorithms for how it works, using eBPF.
Oh, so on the fly, essentially.
Yeah, you don't have to build a custom new kernel module that implements this
queue disk sort of interface. You can now do it with eBPF.
Hmm, I wonder how that could be used for dynamic rebalancing of prioritization and stuff.

(45:41):
Maybe driven by some sort of AI that's watching your network.
Or really, just like, you know, what I would love is some integration with Home
Assistant where it's like, okay, streaming mode, and now the set-top box gets
all the traffic prioritization.
And you hit another button, and now it's like my computer for work gets all the prioritization.
Wouldn't that be amazing to be able to integrate it at the Home Assistant level

(46:02):
and then tie it to like a Z-Wave button?
Yes. I love it.
Okay, good news for virtualization this time around. KVM support on RISC-V,
no longer experimental.
Uh-oh.
Huge news for Intel if you want to do confidential guests.
x86 virtual machine hosts on KVM now support TDX, enabling the use of confidential

(46:23):
guests on Intel processors, something already possible on the AMD side,
but has been in the works for literally years.
Quote, literally years, yeah.
Uh-huh, and includes a large number of patches.
And then nested virtualization support on 64-bit ARM is also now working.
It's disabled by default. It's still kind of new, but that's cool.

(46:44):
So RISC-V and ARM seeing some seriously nice virtualization improvements. Yeah.
And then on a note on the way out here, there's now an AI-written shell script
in the kernel source tree.
Really?
Uh-huh. It's called gitresolve.sh, and it was at least co-created with the use of some LLM technology.

(47:05):
I saw, too, there was some discussion, the kernels coming up with some standards around AI tooling.
Yeah, they're doing a lot of discussion around that, yeah.
Yeah, so they're trying to get that sorted out, too. And it doesn't sound like
the rule is no tooling, but it has to follow certain standards.
As LWN says, not only does it work, this script, but it includes a full set

(47:25):
of self-tests, something that the author noted, with understatement,
is unusual for code found in the kernel's scripts directory.
LLMC said, won't give you a frowny face when asked to generate tests.
I noticed the restraint. There was one feature you didn't mention here.
The power management subsystem has gained rust abstractions for managing the

(47:46):
CPU frequency and operating performance points.
Yeah, a whole bunch of stuff there. Also, new abstractions for core memory management operations.
Yeah, and other drivers.
Yeah, there's a lot of good rust work happening without a ton of fanfare just going on under the hood.
It's incredible. This and this, by the way, this is the high level.
This is us doing like there's all kinds of security improvements and fixes across

(48:12):
all kinds of subsystems and new features that we didn't even touch.
that's where kernel newbies can really do
a heavy lift and if you really want to geek out and
spend an evening reading through everything we'll link to
some of the best resources in our show notes this week so you can do just that
because what other project in the world is kicking ass this hard this far in

(48:33):
and shipping stuff like this on a general platform that's available to the entire
world and everyone can build on top of this massive win by the kernel team 616 is a winner.
Unraid.net slash unplugged. Unleash your hardware and see what you can build.

(48:56):
If you're ready for the ultimate homelab that'll scale with your skills and
your desires as they grow, Unraid is what you have been waiting for.
I know a lot of you are still very interested in virtualization.
Okay, what? No, I tease. Actually, it's still great. And Unraid makes it even
more approachable, not only with a nice interface, but they're sitting on top
of a stack of winning Linux virtualization technologies.

(49:19):
They give you a brilliant interface to pass through things like your GPU or other hardware.
They have support for Virgil support, which means you get virtualized 3D graphics.
There's some potential there when you start thinking about maybe centralizing
one nice graphics card and then using your machines around the network to access
that graphics card, so you only have to buy one, you know?

(49:41):
And then also there's a whole ginormous library of community applications,
including a lot around local LLMs.
So if you want to deploy, there's several different options there and there's
even things that will let you essentially set up an open API-like interface
in front of your local LLM so you can use applications that talk to OpenAI but

(50:02):
they're actually talking to your local LLM running on top of Unraid.
And if you haven't checked out Unraid for a minute, 7.1 is taking things to the next level.
With the ZFS support for one, They still support the mismatched drive sizes,
so you can upgrade as you have capacity as a new drive comes into your rotation,
or if you've got some dis sitting in the closet right now that aren't all the same size.

(50:24):
And they've made a really big improvement as far as I'm concerned.
Built-in wireless networking.
Now, I know that's a weird one for a lot of you, but for me,
where I can't run Ethernet,
Huge. And the other thing that I really appreciate, I think this really kind
of speaks to their whole model, is they're building in the latest Linux features, right?
Faster Linux kernel ships in 7.1, for example.

(50:44):
That's because they have a sustainable model here. They ask a reasonable price
and they continue to build this thing.
They've been doing that for a long time.
and if you go to unraid.net slash unplugged, you can try it for free,
30 days, no credit card required, a great way to support the show and a great
way to scratch that itch, deploy some applications really quickly,
utilize some disks you have sitting around, maybe you've got an old laptop or

(51:05):
an old PC or maybe you want to build something.
Unraid can be that secret little bit that just gets you going fast and you know
you're building a reliable, scalable, long-time supportable system.
So check it out. Go to unraid.net slash unplugged.
Well, we have all sorts of shout outs to make this week. We have a bunch of

(51:29):
new members to Jupyter Party and others.
Nicholas G joined Jupyter Party this week, along with Hoa F.
And then we have a bunch of core contributors to Linux Unplugged.
Matthew Y., Brett B., and Jonathan S. Thank you for joining us in the party.
Yes thank you for supporting us we really

(51:52):
appreciate it hope you enjoy the member content check out the
bootleg there is the bootleg promo code
it's still active it takes 15 off the unplugged core contributor or the whole
network jupiter.party membership and it's going fast there are four slots left
as of this episode i went and double checked the accounting four slots left
for the promo code bootleg to get in there and support the show directly and

(52:15):
get the extra bonus member content,
We have a shout-out, too, for the community this week. Wes, I think you should
take this one because you were there in the thick of it right after I kind of
just created a massive problem out of the blue. Just blew things up.
Yeah. Well, you were trying to, I think, provide some good value for your This

(52:36):
Week in Bitcoin members by putting on some member-exclusive stuff with Fountain's
new Podcasting 2.0 sort of member content.
A little bonus content. A little bonus content. but.
The system required that like it also inserted a trailer in the public feed.
Yeah it's it's neat because it's all feed-based so you don't
have to have like a special platform but that means that

(52:58):
all of a sudden instead of having a full episode in the feed we have like a
one minute trailer and the system is designed well a new item in the feed let
me go automatically create a post on the website and we don't want to post on
the website for a one minute trailer yeah so um sorry everybody sorry i didn't warn you you.
Like to experiment in production uh and uh thankfully uh chance our dear scraper

(53:22):
and website co-maintainer there and scraper maintainer was quick in action to get it all fixed up.
Yeah so we weren't posting silly little one-minute trailers to the jupiter broadcasting
website after i just randomly had one pop up in the feed so.
We are super appreciative of that.
Can i post an episode at all right now yeah.
It's all ready to go.
Man thank you chance boy i really appreciate that.

(53:42):
What i heard you ask there was can i break it again this week.
Will it work will it work.
We do have a couple pieces of mail
from the mail bag here that we'd like to push to the front one the first one
here from aiden hey all i like the discussion on managing nixos configs in the
last episode and wanted to share the method to my madness i keep all my configs

(54:03):
in a flakes repo on github well done which is managing around a dozen hosts.
I keep my user preferences like shells, aliases, GNOME, configuration extensions, etc.
all in a file for my user and each host has its own file to add unique settings like the hostname.
I'm using it to manage my home server as well as several workstations and gaming

(54:26):
setups used by both myself and some of my family members.
It makes pushing out software to computers in the house incredibly easy,
and I can get a new computer added to it with just a short commit and passing
the flake name to NixOS install.
And here's a link to that repo.
This seems like the setup I basically need to get on.

(54:47):
And now you can.
I really should. So...
I have really resisted this kind of thing, but the idea that I could have a
system where maybe I make a few changes in a config file,
commit that to my Git repo, and then my downstream systems just assimilate that
change and adapt maybe even automatically in the background,

(55:07):
and build a new image, and the next time I reboot I just have the changes.
Did I hear you say you're going to start using Git to finally track your Nexo as config?
I really should, because right now I just keep everything in my home directory.
It's easy to get started.
Uh it you know it's like it's like setting up mqtt i just i just don't want

(55:30):
to have to do it i just i want it to be simple enough where i don't have to
you know i love the days where i could take the whole system copy to a floppy
disk and then i could take that to another system and copy it over and i'm done well.
You can do that.
You just have to.
Get a floppy that works.
I'm thinking that's what i should do wouldn't that be funny if i just went hard
the other direction this goes back three.
Separate floppies at all times with your.

(55:50):
Next So that's what I revision. I just, yeah.
You got a little label maker.
The problem is the era of the floppy disk, as we discovered when we did a little
experimenting, is over.
USB floppy disks do not show up as a traditional floppy drive.
They show up as more like a, almost like a CD-ROM or a virtual USB storage device.

(56:10):
I hear tapes are still.
Dude, I love tapes. I'm all about that.
I just want to say Aiden this comes at a perfect time because I stayed up way
too late last night with my brother trying to figure out how to do a multi host
Nixos flake so thank you for saving us dude.
This is like when he repaired the van door and didn't include me it's like bro

(56:34):
this is like what we bro up about and then like he just bros out without us.
I guess with his original bro yeah.
You're right about that well I was probably tossing and turning.
Okay, I'll try next time.
All right, Kmog wrote in, Remember here, re-immutable use using NixOS everywhere at home.
I'm also a Nix packages maintainer for the bucket software. Well,

(56:56):
thank you. That's awesome.
I am looking at perhaps Fedora Core for work.
I think it would work as a great tailscale subnet router. Since those machines
have one job and being Red Hat adjacent, I can get corporate buy-in.
As far as NixOS config version control, I put a really generic file on the machine
that imports from my personal Git server using SSH. Oh, neat.

(57:17):
This way I can work with the configuration from anywhere and simply push the
new config to the server. I really don't care what happens to the local at scenics OS folder.
That's a great setup too. I like that KMogged. And thank you for being a member too.
Okay, I need to up my game. I can feel I'm old manning it here. I can feel it. I can tell.
There's something about it that just makes me want to lean into it.

(57:38):
I am probably required at this point just to note you could also do that dream
you described using a, you know, Bootsy OCI system to push out stuff.
Well, yeah. Yeah, but still, it all comes back to having something in like a central repository.
Yeah, true.
It does. It's just right now it lives in my home directory. Thank you everybody
who emailed us this week, linuxunplugged.com slash contact if you'd like to send us an email.

(58:02):
And we did get some boost to support this show, which is direct support from
our audience. And we start with our baller booster this week.
And it's a baller indeed. A-A-Ron comes in with 100,000 sets.
All right. Thank you very much. That's a fantastic boost.

(58:31):
Made our day when we saw that one come in. Aaron writes, I believe Home Assistant
has a built-in migration tool for the Zigbee coordinators to make it easy to upgrade.
Since you'll have to buy a coordinator if you get new hardware anyway,
maybe get the coordinator first, then migrate to the new coordinator using your
HA Yellow, then theoretically all you would need to do is plug it into the new hardware.

(58:56):
Not so sure about Z-Wave or Matter, but I bet they would be similar.
Anyway, it's been a while since I boosted, so here's a little extra show,
some love. Sad to see self-hosted go, but happy to have some of those memories.
Stay savvy. P.S. What would it take to get my own soundbite? Well, I think...
I think you got it, buddy. I think you got it. That one got you too,

(59:16):
my friend. That one got you too. You keep boosting, we'll keep them on the soundboard.
Thank you so much for that idea. That's totally how I'm going to do it.
Yeah, that sounds like a great way.
That's totally how I'm going to do my migration, Aaron. I really appreciate
that. Thank you very much for the boost.
Turd Ferguson boosts in with 22,222 cents.

(59:42):
Ooh, a bit off topic here. On again, off again, Star Trek fan, but I know y'all go deep.
What are your thoughts on the new Star Trek Academy series?
It's never off topic, Wes, although we'll try to keep it brief for those that
don't care. I would sum it up as a Star Trek series nobody asked for.
Brutal but right I think.

(01:00:03):
But I think if you enjoyed Discovery you're probably going to enjoy this same universe it sure looks.
Similar from the like.
You know the production we're going to watch it oh yeah we're going to watch it which Star.
Trek have you not watched section 31.
Yeah if we watch section 31 we're going to watch Academy right it's going to happen so I.
Don't know if anyone else wants to boost it.

(01:00:24):
Thank you turd appreciate the boost Well.
Derivation dingus comes in with 20,000 sets.
Well, I almost tried Nebula, but I actually just converted all my stuff from
WireGuard to a self-hosted NetBird.
If you guys had consulted with me first, I'd have probably chosen Nebula,

(01:00:46):
but at least it'll be interesting to compare the two.
With WireGuard, I kept ending up in situations where DNS alone couldn't quite do everything I needed.
NetBird redirects DNS queries to itself, which allows it to respond to your
private network or for them to the local DNS provider.
and that means I can refer to my stuff by name and it doesn't break the software

(01:01:09):
on my work PC. What a time to be alive.
Woo!
I really like it so far. It's much better than handling as many WireGuard keys as I was previously.
Does Nebula do something similar to NetBird with DNS?
I wonder, Wes, could you do what I do with my TailScale network is I run now
a DNS server in my TailNet.

(01:01:29):
Could I, like, on one of my lighthouses, just run a DNS server?
Hey thanks for telling us about your setup uh self-hosted netbird
is a great option as well for sure the nebula dns
is newer and does not do as fancy stuff so like netbird acts as the dns and
then can kind of redirect things nebula only responds right now to things for
that it knows about so it works better in the situation you're describing where

(01:01:51):
you have a an upstream dns that can delegate and return responses you know that
it learns about a proxy from nebula having.
Tried both systems, I prefer one where I have slightly more control.
I do kind of end up running a DNS server like that anyway, so it's nice when
it's incorporated. Yeah.
Yeah. But I love hearing about people's setups, so keep sending them in.
And thanks, derivation. It's good to hear from you.

(01:02:14):
Fuzzy Mistboard is here with 4,444 SATs.
This is a great episode. I enjoyed the self-hosted discussion and some interesting mini hardware.
I just made the switch to LinkWard in about a week before you found it.
And by the way, it's great. I've pulled all the Z-Wave and Zigbee manuals for
my devices as well as manuals for my motherboards.
Oh, that's such a great idea. And all the other hardware I want to preserve.

(01:02:37):
LinkWard made it super easy to pull them in, tag, and organize.
Okay, that's an excellent experience report. I haven't set it up yet,
but it's really somewhere on that to-do list.
Why did I not think of this? Like, literally, so my wife just got a juicer.
And I'm like, well, what do we do with these manuals? I can probably find PDFs, but do I keep them?
And she's like, well, I really only want the recipe book. You know,

(01:02:58):
it's like a system that just is designed to pull in PDFs and organize and tag them.
That's what I need. That's great, Fuzzy. Thank you.
I got to really get it going for her.
Outdoor Geek boosts in with 5,000 sets.
this zigbee dongle linked contains the same soc as the ha yellow oh sonoff zigbee 3.0 usb dongle.

(01:03:25):
Well well well at 17 you guys are coming through thank you.
For migration i would make sure the firmware versions match firmware can be
flashed i bought this dongle a year ago but only just plugged it into my Pi 4 home assistant.
I've seen some chatter about bad knockoffs, so try and buy from a reputable source.

(01:03:46):
Outdoor geek, that's very, very valuable. Thank you.
We have some more boosts here. One from MG1010104444s.
A couple here just saying setting up AlbiHub.
Hey, right on. That's the self-hosted route to go.

(01:04:09):
AlbiHub just had a new release, and it is even easier to manage than ever now.
If you want to dip your toes, it's a technical deep dive that I think you'll find very rewarding.
Well done, MG. Thanks for the boost.
Not a zip code comes in with 4,321 sats.
That's right. On the little red dot, just the zip code tells you where to go, just on the little dot.

(01:04:35):
So I'm not boosting a zip code, but nonetheless, looking forward to episode
700 to see how many local listeners we have.
Or if you're in Singapore, please boost in.
Okay, so this is a response to our not a zip code from last week.
I think I said Singapore.
Yeah.
I think I did.
Somehow, I guess we got it right. Because this was where it was like per building,
right? The code told you the building?

(01:04:56):
Yes. I've been wanting to visit Singapore for some reason. I don't know.
I have a buddy who went there and just raved about it.
And now ever since, it was like two years ago, ever since I've been.
Singapore meetup.
Oh, my gosh. Also, we should have a hell of a party for 700.
700. We should have a hell of a party.
Someone do the math of when that lands.

(01:05:17):
Oof. Thanks. Appreciate that. Not a zip.
Caffeinated Linux comes in with 4,242 sats, 4242.
Boosting in with my immutable experience, I've been using Fedora Silverblue
on everything but my main workstation for a few years now. It's been a great experience.

(01:05:39):
Smooth updates, Flatpak apps work great, and I only have a few packages overlaid for some drivers.
I was even able to rebase a spare laptop from Silverblue to Cosmic Atomic using
one command and after a reboot i was ready to go in cosmic love the show keep
up the great work p.s uh all 4242 sats earned by listening on fountain.

(01:06:00):
Very nice it's impressive um i also have rebased a couple of times i've rebased
three times now and it's worked great i really love that it's a it's something
that's just so slick about this experience and it's it makes distro hopping seem old.
Because you can rebase to an entire plasma system, cosmic system,
gnome system, but your user data and your applications and everything the way

(01:06:23):
you have that configured remains.
No reinstall or super careful partitioning or anything like that.
Well, a dude who's just trying stuff tried to send us a row of ducks and succeeded.
Yes, this is the Homelab content we were promised.
I'm one of those who dearly missed the self-hosted show, and I'm glad it is

(01:06:43):
living on through nods to self-hosting in Linux Unplugged. Keep being awesome, lads.
Well, thank you.
P.S. I'd love if more of the picks were useful. Self-hosted apps.
The way you said that.
Quit with the two-y crap and give us self-hosted picks.
Yeah, just the way you said that. It would be great if they're useful.
We only do un-useful self-hosted apps.

(01:07:05):
Okay, so I don't have a bunch of self-hosted ones, but I do think they're going
to be useful. But I agree.
More and more, I've been like, well, you know, we don't need a whole segment
on this, but we'll definitely put this in the pick. And a lot of them are some
of the self-hosted pick stuff.
Yeah, that's a good idea.
That community, it kind of blooms lots of good stuff, and then it kind of goes
quiet for a little bit, and then lots of new stuff blooms. So it's really fun to watch.

(01:07:26):
Thank you, everybody, who supported the show. Aaron, you did a heavy lift this week.
Thank you very much. And thank you, everybody, who streamed.
We had 24 of you stream as you listen, stacking 18,439 sats collectively.
Not our biggest number there, but I appreciate that. When you combine that with
our boost, the show eked out 185,545 sats.

(01:07:50):
Those go directly to the host, to editor Drew, and to the podcast creator and the podcast index.
The splits are defined in the RSS feed, so you can essentially see the contract
in XML, and it goes directly to everybody involved. It's just a great way to
support each individual production and get your message on here.
Fountain.fm makes it really easy to get started, and they're about to release

(01:08:11):
a banger of an update that'll include ways to boost even easier around the world.
But of course, you can go the self-hosted route with AlbiHub or skip all of
that by becoming a member and putting your support on autopilot.
linuxunplugged.com slash membership for this show or jupiter.party for the whole network.

(01:08:48):
Now we do have some of those picks to get to before we get out of here.
And this first one is Tony.
Tooney? I don't know.
I don't know. Tony?
Damn it. I hate that we have to do this so often.
Go with Tooney. Come on.
It's not this hard in the commercial world. It really isn't.
because marketing forces them to use words people know.
But it's a fast, lightweight terminal, note-taking app.

(01:09:09):
Yeah, that's right. How about...
Sleek TUI interface built with Bubble T and Go, which is a great library for this kind of thing.
Markdown rendering right there, file navigation, and native NeoVim editing right in your terminal.
Pretty nice, Wes. Pretty nice. And then on the other side of the spectrum, oh, MIT licensed.

(01:09:30):
On the other side of the spectrum, we have something built around Etsy Sync.
This may ring a bell. It was recently, semi-recently open sourced.
It's interesting encrypted synchronization technology.
And Etsy Sync KNotes is built on top of that encrypted syncing technology.
And the Notes application uses a connection to one of your Etsy Sync accounts.

(01:09:51):
It can be self-hosted or one of their hosted ones. And you can create notes,
edit notes, search across your notes, sort the notes, display notes,
all the things you expect with Notes.
But the nice feature is it's actually an application taking advantage of Etsy
Sync. and if that's something you've been following or interested to see somebody
develop software around, we're actually starting to see it now.

(01:10:12):
That was the big thing I was waiting to see. I mean, there's a couple of core
applications they already have and they're really cool, but this Notes app,
it's lean, it's mean, but it's taking advantage of this pretty powerful sync technology.
I feel like sometimes with the picks, we can tell maybe some themes on what
is happening behind the scenes at JB.
So are one of you looking for a new Notes app? Is that what's going on here?

(01:10:36):
I'm thinking about it. I'm still living the Obsidian lifestyle,
but I find that what I tend to do is find a leaner, meaner app to just create
my notes, and then later on I sit down and copy and paste it into Obsidian, and it's got to go.
You know that's wild, right? Why would you ever do that?
It's got to go.
So you need a plug-in that auto-imports your...
There you go. There you go. Yeah. If anybody has some suggestions for something

(01:10:59):
that works on desktop, mobile, iOS, Android, and syncs,
i mean and i did not evernote not obsidian you know that's why i was looking
at etsy sync i don't think it's there for me obsidian is a hard one to replace
you know because those community plugins i love those sweet sweet community plugins can.
You remind us what it is about obsidian that has you looking elsewhere.

(01:11:20):
Well i think i'm parted so i pay so much per year for the sync and then it feels
kind of like a heavy application so i don't load it very much,
if i didn't use any ios devices it'd be fine but
when i'm working on the car i like to use the ipad for taking notes
but then when i don't have the ipad i end up using like the garage computer
or my phone and i just like open up a text editor and just take the notes on

(01:11:42):
my phone it's just it's just not working for me something has to be on the phone
i you know i don't know and i've tried next cloud notes too maybe it's me i think the problem is me,
I'm not quite willing to give up yet, you know? And I can't just give in to, what is it that you love?
Logseek.
Yeah, Logseek.
I think I have a new beta mobile app, actually, I need to try.

(01:12:04):
Yeah, you and your... I said to overstruggle, meanwhile, you're just over there
loving your Logseek. Drives me crazy.
I'm not going to say it's perfect, but I do like it.
And Brent's just using pen and paper, right?
It's strapped to my leg.
Yeah, but he makes the paper himself.
He does.
From Canadian trees.
Oh, I thought it was toilet paper.
Oh, okay. Canadian toilet paper.
All right, well, let us know what you thought about the FlatHub-only desktop

(01:12:26):
app future, and particularly for you.
What tools, what workflow would break for you?
And if app packaging did stop being a differentiator for distributions,
what would make you pick?
Fedora over Debian, or whatever the district might be, say, in a few years,
if this future were to materialize.
And why is it Arch Linux?

(01:12:48):
Wes, leave him with a hot tip before we go. Give him a little more you know.
Oh, yeah. Well, you might not know, but you should. We're a podcasting 2.0 podcast,
which means we have chapters so you can skip right to the content you like and
transcripts. Search it.
Use it to display. Follow along. Skip even finer detail.
Yeah, we try to link everything we mentioned, but maybe we mentioned something

(01:13:10):
we didn't link. You can probably get the name of it or close to it in that transcript.
That can be pretty useful.
See, a little listener pro tip for you right there at the end.
Also, one last pro tip. We're live.
Go ahead and make it a Tuesday on a Sunday over at jblive.tv or jblive.fm.

(01:13:30):
Kick it up a notch, grab yourself a beverage, and join our mumble room.
Our virtual lug gets together around 10 a.m. Pacific.
You can get it over at jupyterbroadcasting.com slash calendar in your time and
our mumble server info that's over at jupyterbroadcasting.com slash mumble.
Now, links to everything we talked about today, well we try to put all of that
at linuxunplugged.com slash 625.

(01:13:52):
You'll find other links like our contact form, the RSS feed,
a bunch of other things over there, like the membership as well over at the
linuxunplugged.com website.
Thank you so much for joining us on this week's episode of the Unplugged program.
We'll see you right back here next Sunday.

(01:14:57):
Uh breaking checking in from the past news someone just checked in from the
past yeah uh simulated snakes checked in via matrix checking in from the past
um we're not calling it the bang bus anymore are we.
Oh yes yes yes i love the time travel check-ins thank you everybody who continues
to do that just if you're listening in the past now you know weeks or months

(01:15:20):
after this episode came out try to check in in real time with us and let us
know where in the timeline you We'll be listening from the future.

All Episodes

Episode Transcript

Popular Podcasts

Stuff You Should Know

The Joe Rogan Experience

Dateline NBC

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}625: They're Doing it Wrong!

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Stuff You Should Know

The Joe Rogan Experience

Dateline NBC

All Episodes

625: They're Doing it Wrong!

Stuff You Should Know