Hello and welcome to episode 74 for Linux down time, I'm Joe, I'm Gary, I'm Amalith,
I'm Faelham and I'm George. Good to talk to you all. Welcome back George and welcome
Faelham for the first time to this particular show. People may know you from late night
Linux. Howdy. So, Faelham and George on one show this can only mean one thing, you're
gonna have a row. Are we arguing? We are arguing yes. Yeah, you're gonna have a row about
a mutable file system distros. Now, what's gonna hopefully happen here is Faelham is gonna
outline his concerns and George, you're gonna allay those concerns and fears and Faelham
will be installing you blue by the end of the recording. I'll have you know, I even fired
up kinwise for the last day just to see what it was all about again, because I couldn't
remember it from last time. We're ever pronouncing it right now. We're definitely improving.
I did geology. It's all good. Right. So, Faelham, what are your main concerns with this type
of Linux desktop? I guess they are concerns. They're not very staunch. You know, I'm not worried
for the future massively, but I do have concerns that we're gonna turn a nice flexible adaptable
Linux, which is can be as small as you want it into a monolithic image based have it in one lump
need to restart every time there's a single file update, you know, nano gets a patch reboot
your machine to take advantage of it. We're trying to drive every towards this way of doing things
for very much in quotes security and consistency when I'm not entirely convinced about the security
side of it. It'll be consistent, all right, but I have never struggled with inconsistent distros.
Maybe I've just been lucky, but it seems like a very overly complicated way and a very hacky way
to do this on top of the current system, where I think so far we've done really well. In the fact
that you can look behind the curtain on a Linux distro, you can with just a bit of poking see things.
This isn't like OS 10, MacOS, whatever it's called now, or Windows where you take what you're
given, you can't touch it and you don't get to play in the background, whereas we do and always have
and I think that has been our sort of superpower is the fact that people have been able to do that.
Wow, there's a lot there. Yeah. So where are we starting? You've got 30 seconds go. Are we starting
with the flexibility? So I get it, I understand. The thing I usually try to tell people is you're doing
everything that you're doing now. You're just adding an extra stamping process at the end. So
if you're using Linux now and you're totally happy with it, I'm not a threat to you.
You see, I don't buy that, right? Because I'm pretty sure when linear was going on and on about
system D and how well, if you don't like it, do your own thing and well, yeah, you can use dev one.
Oh, cool. Yeah, me and all the 10 guys go over and use that. Yeah, it'll be great. And we'll get
it all tested and it'll be brilliant. Whereas I think reality is if something with the backing
red hat and some other players rolls that way, pretty much everything goes and rolls with it. And
you know, that's just tough. It's de facto standard is also a standard. So are we talking about
immutable Linux? Is there we talk about system D because that did solve a lot of problems. Yeah,
well, no, I'm using system D as the example. I mean, I like system D. It's fine. I know some people
didn't, but try to avoid it now. And I think you're facing a feral pill struggle or a very small
niche distro. I would agree with that. Yes. But I did what I worry is that you saying that
immutable is not a threat. Well, immutable might go that way too. And then it's a case of, oh,
yeah, there's me and you know, I'm a little me on make our own distro and you can't play.
Well, old distros don't go away. So Fedora will still be there. There's the methodology, the packages,
all the processes that your distro, whatever distro you use uses, they don't go away. If Fedora says,
let's make the immutable the main one that we link to on the website. The RPM packages don't
not show up that day for work. I mean, not yet, but you know, give that six months, a year,
two years, and they're gone. Yeah, but what would you replace them with? It's an additional step
at the end. And I think a lot of the processes that are used to make distros are reusable. We're
just kind of stamping them out at the end. I would probably make the same argument you can make for
system D for all of the loud minority that is complaining about system D. How many people have
actually had to debug someone else's homemade bash system? It's script. I haven't had to do that
in 15 odd years, wherever long it's been like the industry decided that system D was going to be
the standard. And some people don't like that. And that's your choice. But if you show up to work,
and you say, I'm going to rip out all the production because I want to replace the admit demon,
like people don't even have those conversations at work. Do they? I hope not.
So I think that when you have a problem that is solved in a certain opinionated way, it does lead
to people kind of being wary about it. One of the things that does attract me to it is that it is
being well-funded by Red Hat. And they're using, in this case, OS tree. It's well-funded. They're
going to use it for OpenShift. And I'm going to use me as a cloud native person. I'm going to use
that to my advantage to take advantage of that thing that I can use for free. However, if you could
say, well, I don't really like that. The composability of RPM OS trees too complicated. Or I just want a
more normal looking thing. You can also just grab OpenSusa micro OS that looks like OpenSusa. It
feels like OpenSusa. And it's just using file system snapshots. So it's similarly ish to what
traditional distros are doing. But if you don't want any of that, you know, George says I have to set
up a CICD pipeline now. And you know, make my own image, you make my own custom image, things like
that. I don't think advanced users that want a little bit of customization would be interested in
things like that. But I think what we're seeing right now and what we're seeing a little bit in
Universal Blue is people start to kind of congregate around things that are important to them.
So right now, there is no one person that's saying, I'm going to make the smallest tiling window
manager. You mentioned the example of a really small image. All that takes is someone saying, okay,
I'm going to make a really small image. Right now, the focus at least that I'm seeing is a lot
of people are actually trying to add things to their image because it removes so much more of the
complexity that people are kind of willing to make that sacrifice. But to counter that though,
if you want to make a really small installation where the tiling window manager or something with
Fedora, that is just a few DNF commands at the moment. You don't need to know about building an
image. Yeah, for sure. I'm not saying that all Linux users need to learn how to make custom
images. And instead of right clicking on your desktop, you're going to have to go into GitHub to
get them installed on your image. I think what's going to happen and what is happening is little
communities will be forming around these images and you're just going to pick one like how you
pick a distro now. Yeah, but do we not end up diluting ourselves more than I mean, it seems like
we were almost getting to the point where we had some good strong candidates. We had a very reliable
system. Steam is now doing well on Linux. And now we're going, well, why don't we split this all
up? Is it reliable? Well, yeah, I think it is reliable. I don't have issues with my machine.
How long have you been using Linux? Since, I don't know, late 90s. But I use it as a daily driver.
I do my work in it. I don't use any other OS. And if I'm going to do risky stuff,
I'll fire up a VM because I know a VM can be literally blown away and not matter. I know if I'm
going to run into a dungeon, it's really contained. It's not hopefully going to be containerized
in some way. It's almost as close to a physical box that it can be in. So I tried to kin away
because I haven't used it in so long that I said, okay, I need to try it again. And there was
an update to the image for several packages. And I would have thought, well, surely it should
be so small that it's only system stuff that's getting updated in the image. And maybe that's
where you're talking about, maybe it's more work if they do it that way. But I literally installed
a few packages and it said I need to reboot. This just seems like a crazy situation where we've
got this reliable system. We can have thousands of errors of uptime. And now we've got like
a nano is getting an update and I need to now reboot my VC to take advantage of it. It's just weird.
Whereas I would have thought it would have been far more integrate with say flat packs where you
would have as much stuff in flat pack as possible. But it doesn't seem to go that way.
Usually my image I treat as one whole not as an individual package. So the things on my image,
it's like a basic GNOME desktop with some extra bling on it. Fedora does update a lot. So that day,
it might be nano, it might be G-lib C, it might be the kernel, it might be those other things
on that image. But that's generally transparent to me. I don't sit there and say time to reboot.
I have a new nano. I just kind of use my computer and then the updates just come to me.
The problem I would have is the fact that you have to do that reboot. That's a huge break in time.
Like right now, I have a bug with Katie that I have to log out and back into my session. I think
it's a massive bug. You think, oh, log out and back in, no big deal. It's a massive deal.
Yeah, losing your session sucks for sure. I've got all the SSH connections to all the different
clients to the fire backup, reconnect to the VM machines. They're still running, obviously,
but it's a mess and it takes so much out your day where you go, oh, Christ, need to do that now,
because you're getting black windows, whatever. That's a huge thing. If I look at the
Kino, I set up, I've got here, there's two images. There's the main image and the fallback image.
And I was like, well, where's Firefox then? And I look for Firefox. Firefox is in that image.
I would have expected to see Firefox would be a flatback. It is not. Firefox 112 is inside that image.
This is like windows. The worst thing about windows is doing updates and windows because they're just
a morphous massive blobs that come in and we're lucky that these come down quick. But I don't
want to be taking huge massive chunks of data down just for one tiny thing. It's bizarre.
So this is an area that I actually disagree with what Fedor is doing. I don't think browsers
should be on base images. And on my personal, in my Bluffin image, I remove that.
For Universal Blues, the initial images, the main and Nvidia images, we try not to
editorialize too much of what Fedor is trying to do. The first cut actually, I ripped out all that
stuff. I was like, I'm going to fix this thing. I'm going to fix this distro and take it that last
10%. And some people like that, but I also got a lot of people who were like, look, I just want
Codex. Don't make the editorial decisions for me. So us as a project, we kind of looked
and said, well, I really hate that RPM Firefox. It is not good. But you also kind of have to respect
Fedor's decisions there. Well, here's an interesting idea. Is it because they feel that flatback is
in some way deficient for some of the packages, maybe. I'll give you an example. It's a snap-based
example. And I had species of species that flatback is similar. I was at a client's office and they
said, open office is not open and stuff off the desktop properly. I was like, right? What do you
mean? And so they fired it up. It was actually turned out to be a sambar share off to a Windows box.
And snap wouldn't access that. You had to copy it for locally and open it that way. I thought, what?
These are the kind of issues that going from an OS that where everything is everywhere to
everything is going to be sandbox Android style where it's like you have to give things
explicit permission and things like that. So those are just transitionary things that still need
to be solved. I know I happen to have two win-gie examples that happen to. Oh, no,
the time mean people have these. Yeah. It does not make the system sort of, I don't want to say
shoots holes in it, but if there's not many problems with it, you're not going to wake up one day
and it's going to be like all the flat-pack apps now just magically work with your sambar shares.
This is a model transition and the problem that needs to be fixed is applications and certain
intersections of those applications with what user expectations are do not fit the current
model that we're trying to go for. Like if phones had already existed and then you introduced
Android, you'd have the same problem except Android kind of came that way. When you're on the Android
phone, is there an app that lets you just browse shares? Father browser does it for me. Certainly
the one I installed from after I does. I don't know what you can see everything with it, mind you.
Yeah, I'm just curious because a Linux out there with this model has solved that. So my question
is kind of what does the UX look like? Does it ask you for permission? Like if you had a tablet
and you brought it in there and you wanted to browse the share. To be honest, I'm not sure. I mean,
before you came on, me and I both were bitching about exactly Android, so I let him share about
this one. My problem with immutable file system distors is that they seem to be striving to replicate
the Android model of interacting with your system. Everything's in a container. Those applications
don't have access to your file system unless you explicitly give that permission and things
like that. I really don't want that UX paradigm on my desktop. There's so much friction just
trying to use my device in, I guess, more of an advanced way as like a developer, this admin type
person that I am. I would not want my desktop to be like my phone with all these pop-ups and
asking for permission and not having the concept of a file system, having to share a file from
one app to another. It's just a pain, honestly. I don't really have any answer to that. I mean,
it's all your fault. That's what it is. The ones that get me are always the day zero ones
definitely get you. So if you buy a new SSD for your steam, you need a new two terabyte drive
because games are crazy now, right? You amount it to FS tab. It doesn't show up in the steam flat pack
at all in the window. You have to explicitly set an override. That's really annoying. However,
my setting up a new disk in Linux checklist went from like three items to four items and that's
just something that I've gotten used to and I've just documented for myself. So I get a new disk.
I have to make a system deune. I'm sorry. Do you all mangle your FS tabs or do you make
system de-mount units? I'm just curious. FS tab. FS tab. Really all of you? Yeah. Wow. Awesome.
Okay. So I have to make a system de-mount unit because I like to have one file per drive because I
keep all that and get. And then I add the flat pack override and it just becomes another thing.
The problem that I think needs to be solved as a whole is adding a new disk in Linux sucks for end
users, including editing the FS tab, setting the flat pack override. And there's no gooey. Like,
the best thing is that the one that I've used that can know disk editor thing. I'm not asking my
dad to use that because that's not going to work for him. So in a way, I think that the problem
that needs to be solved is adding extra disks is not user friendly. That's fair. But you could also
say we could go back to the days of the format words of the disks, right? The preformatted floppy
is where it was. This one's Windows. This one's Mac. Maybe we just need to get hard drives that come
with Linux, right? And it's as extended for. There you go. There's half your problem solved.
But then you'd have an argument about whether it's XFS or XT4. Yeah.
But we're still not. Still not. Okay. This episode is sponsored by Factor. Now that it's summer,
you might be looking for wholesome, convenient meals for sunny, active days. Factor can help you
feel up fast with flavorful and nutritious ready-to-eat meals delivered straight to your door.
You'll save time, eat well and stay on track reaching your goals. With Factor, skip the trip
to the grocery store, skip chopping, preparing and cleaning up too, while still getting the
flavor and nutritional quality you need. Factor's fresh, never frozen meals are ready in just two
minutes, so all you have to do is heat them up and enjoy. Looking for calorie conscious options
this summer? Try delicious dietitian approved calorie smart meals with around or less than
550 calories per serving. My two and a half admins co-host Jim tried Factor and said the meals
were quick and easy to prepare and liked that there was plenty of variety. So support the show
and go to FactorMales.com slash LDT50 and use code LDT50 to get 50% off your first box.
Quick bit of admin then. First of all, thank you everyone who supports us with
PayPal and Patreon. We really do appreciate that. And remember that for various amounts on Patreon,
you can get an advert free RSS feed of either just this show or all the shows in the late-night
Linux family. You can get a link to the Patreon at linuxdowntime.com slash support.
And if you want to get in contact with us, you can email show at linuxdowntime.com.
You've kind of got into a lot of the details and the small problems that you have failed in.
And you've addressed them to some extent, George. Well, try and the best I can.
Debatable. Right, but I think that failim's main problem stems from what you said very early
on George and that was that you are a cloud native person and you are applying cloud native
principles to the desktop. It's built with CICD systems. And failim is an old school sysadmin.
Careful now. Well, we've talked many times about the cloud versus on-prem and all that.
Sure, but I mean, I'm not against automation. I use Salt Stack to the max as best I can.
And I really hate that phrase servers as cattle not pets. It's ridiculous. I hate that one.
It's just absolutely the most condescending statement ever.
You can have a very prized fucking machine that is not a pet.
It's just happens to run a very expensive piece of equipment that you may only have one of.
It's still a server. A server is a server. I love my pet. I love my beagle.
There's servers out there load bearing servers that are pets that if they don't work,
I'm sure the modern world will stop working. When I say those terms, what I try to convey,
hopefully it doesn't come off as condescending is does.
Hopefully I can explain why I'm trying to not kind of sound condescending is. For most cases,
when it comes to running infrastructure, you want to be able to make Budweiser, not MicroBruy.
I love beer, but sometimes you need that mass market scalability.
That's the worst analogy ever. Why? Because Budweiser is produced after you've done
a craft beer and it comes out the other end. That's why.
No, but to return to my point here, the idea of it being built in the cloud with cloud
native technologies, I think that that is a sort of inherent bias that you have against it,
like you just inherently don't like that idea. As a result, you kind of look for things to
nitpick about it. I don't know what that's fair enough. I mean, I'd be okay with CIDC systems,
whatever that's okay, but the problem I would take away is the fact that it's limiting
inflexibility. Fair enough, there might not be like this, but it's my experience with what I've tried.
They're very sort of one big blob. This is what you get, and everything from that point onwards
is you're either fighting that blob or you're layering something on top of it, or I don't know,
I've never tried it with Kino right now. What happens if I install a flat back of Firefox?
Where's that going to end up? Are there going to be 15 Firefox installs in there if I have
14 other users use my PC? That's stuff like that. It seems like a lot of duplication. It's supposed
to be more secure, but I mean, how many strewn about copies of things is going to end up in all
these containers all about the place? That's what I don't fully trust Docker Imgisider, because I think
that's just a lazy way for a developer to not integrate properly into the system and just go,
oh, here's a big box, it works. It's like, yeah, it loads developers. They always say it works,
and that's all they care about, but they don't care about patching stuff or doing all the hard work
of keeping us in running. I worry that that's what we're doing here. We're getting into this whole
the box works. If it doesn't work, throw it away and build a new box. Oh, yeah, it's brilliant,
but we don't go into a nitty gritty of why it didn't work in first place.
Well, yeah, I think the problem that you have here is that the cloud is inherently
disposable and replaceable commodity. That's why I said bow wiser, you don't care.
Yeah, waste product. But your desktop is the ultimate pet arguably. You tune it to be exactly
how you want it. And maybe that is the big problem you have with it, if I didn't.
It's not a pet. It's a tool. And as soon as it starts acting couldly, I'll be taking it out back.
If you're a seasoned system administrator, I'm not trying to convince you to switch to this.
Well, I'm not convinced. We know how to Linux. I'm not planning on hanging out and you being like,
hey, I can't exit out of VR. Can you help me out? Yeah, obviously kill it from another shell.
Like I know, like we talk about reliability and things like that, I can run an Ubuntu system
and it would be totally reliable for whoever I'm running it for. I can be the sys admin.
What I am hoping people like us start to think about is when we realize that we are unique
because this is what you do. You're like a Linux nerd. This is your job.
That doesn't mean that I want to do that when I give a computer to my dad. So
the way I think about it when I'm making my image or whatever is this is something I can give
someone that I know will read some dumb thing on the internet and they will try it and it will
not work on their computer or there was a skew in the a video driver version that day
or something because sometimes people they don't have that choice of their hardware.
I wish I could get a reliable thing. Well, I have to buy you a new laptop. You're not going to
get that stupid optimist thing to work right. Just leave windows on it. I don't know how many of you
have ever had to make that decision. I actually got optimist working and it works pretty well now.
Oh, okay. I think there are two sort of paradigms here that are quote unquote competing.
One of them is treating your desktop like cat only other is treating your desktop like a pet.
I want to relate this to a different industry making guitars. You've got the massive
manufacturers that just pump out the guitars as quickly as possible. They're not hand crafting
their own tools to make those guitars. They're buying off the shelf tools to make those guitars
because they just need to pump them out as quickly as possible. That is a successful industry.
On the other side of the coin, you've got the Luthier who's been hand crafting guitars for 30
years down the road and him hand makes all of his own tools because they're uniquely suited for
his purpose. Both of those industries are successful and valid. They coexist just fine.
You've got some people who want the cheap thing and you've got some people who want the
expensive thing. Both approaches get business. I think as long as one or the other desktop
paradigm doesn't go away, I think they're both perfectly fine, uniquely suited for different
people who have different workflows. Right. Well, that is a very diplomatic way to end. I'm
thank you for that. So we better get out of here. If you've got any thoughts on this,
then you can email show at the nextowntime.com. But we'll be back in a couple of weeks. Until then,
I've been Joe. I've been Gary. I've been I'm Luth. I've been Faelham. And I've been George.
So later.