SoylentNews is people
SoylentNews
Phoronix has an article up about some interesting ideas of Lennart Poettering about what could be a possible future for Linux:
Lennart Poettering of systemd and PulseAudio fame has published a lengthy blog post that shares his vision for how he wishes to change how Linux software systems are put together to address a wide variety of issues. The Btrfs file-system and systemd play big roles with his new vision. Long story short, Lennart is trying to tackle how Linux distributions and software systems themselves are assembled to improve security, deal with the challenges of upstream software vendors integrating into many different distributions, and "the classic Linux distribution scheme is frequently not what end users want."
This discussion has been archived.
No new comments can be posted.
New Linux Vision That Involves btrfs
|
Log In/Create an Account
| Top
| 99 comments
| Search Discussion
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
Remember Darwin; building a better mousetrap merely results in smarter mice.
(Score: 5, Interesting) by Marand on Monday September 01 2014, @10:57PM
This is Lennart Poettering and Kay Sievers. Complaining about historical "problems", deciding they have the only valid solution, and then completely abandoning everything done previously, including any standards that happen to get in their way, while refusing to cooperate with anybody outside of their inner circle (GNOME/RedHat), is on par for these guys. So is letting everybody else scramble to fix the mess after everyone's eventually bullied into doing it through sheer stubbornness and refusal to cooperate with others.
Above statement applies here, too: lack of experience or understanding hasn't stopped them from rewriting everything before, why would it stop them now?
This is what it ultimately is. Poettering is a RedHat employee, GNOME is a RedHat project, systemd is a RedHat project, etc. It's been GNOME's plan -- as stated by GNOME devs -- to create "GNOME as the OS", where the entire software stack is GNOME from top to bottom. GNOME for the desktop, with GNOME-flavoured services and GNOME-flavoured init, on top of a Linux kernel (presumably because they haven't figured out how to rewrite it yet).
As far as they're concerned, any differences in distros are irrelevant because the intent is to control the entire stack and eliminate distros altogether, because nobody should want to use anything but GNOME anyway. Poettering is just saying the same thing GNOME people have been saying for years now. Any distro that isn't RedHat-derived is an obstacle to their GNOME-only vision. You comparing it to "embrace, extend, extinguish" is spot-on, because that's what it is, and other distros are happily marching toward that cliff because GNOME told them so.
It's even more insane when you consider that there are probably fewer distros than ever pushing GNOME as the primary desktop environment because of problems with Gtk3, GNOME3, and the GNOME vision. Debian's enjoying a bit of internal conflict over it, because the GNOME maintainers are pushing hard to get GNOME reinstated as the default. I've seen everything from "GNOME should be the default desktop because of systemd" to "systemd should be the default init because of GNOME".
It's all about controlling the entire stack and strangling the other distros and desktop environments.
--
Also, I liked the bit about "The classic Linux distribution scheme is frequently not what end users want, either. Many users are used to app markets like Android, Windows or iOS/Mac have."
This is especially hilarious because, to an end-user, a Linux distribution is fundamentally indistinguishable from those markets. You have a central repository, all your software is there, and you pick-and-choose without going to random sites to find it. Linux distributions created that model. The premise and conclusion don't even match. You can already add extra repositories if you need to get outside the curated "market", and some distros (such as Ubuntu) provide tools to make this even easier. Debian and derivatives have gotten this right for many years now. You can even, with some care, cross-install across derivatives, such as using Ubuntu PPAs on Debian.
The problem is that end-users don't care about RedHat distros because they didn't get this right before Debian-and-children did, and that hurts the "GNOME+RedHat world domination" plan.
If anybody thinks that isn't the point, consider this: if the distro is just namespaces and filenames, and their goal is just to satisfy end-users, then why doesn't RedHat just pack up shop and consolidate everything into Debian? Debian and its children have a much greater user base than RedHat, and most third-party software already provides .deb packages, so what benefit is there of handing stewardship of this fictional The One Distribution to RedHat instead?
(Score: 4, Insightful) by frojack on Monday September 01 2014, @11:47PM
Still, I hate to see BTRFS getting tared with the brush reserved for SystemD.
BTRFS is stable, reliable and has a lot going for it. Its a Btree-based file system similar to ResiserFS but significantly improved performance and reliability and built in snapshot capability. Opensuse is moving toward it as the default in future releases.
No, you are mistaken. I've always had this sig.
(Score: 1, Insightful) by Anonymous Coward on Monday September 01 2014, @11:57PM
Fair point. If I were involved in BTRFS, I'd be furious enough to puke blood. We can probably expect Lennart to fork it and integrate an incompatible branch right into systemd.
(Score: 0) by Anonymous Coward on Tuesday September 02 2014, @01:18AM
The btrfs developers need to go on the offensive now. They need to make it publically clear that they do not support systemd, that they do not support anyone advocating for the use of systemd, and that they are not in favor of any sort of integration between systemd and btrfs. They need to put an end to these shenanigans before they even begin, solely for the good of btrfs.
(Score: 3, Insightful) by Marand on Tuesday September 02 2014, @12:03AM
I think that's inevitable considering it was an Oracle project in the first place. That's going to damn it in the opinion of a lot of people no matter what its merits may be. Poettering advocating it is small-time compared to the hate people have for Oracle.
Personally, I have no opinion on btrfs, because I'm extremely conservative with my use of filesystems; I have a tendency to hit obscure bugs in software due to niche use cases, and I'd rather not tempt fate with immature filesystems. I'm still using ext3 and XFS, though I'll be switching to ext4 for the next update I do. By the time btrfs is on my radar, they'll be working on btrfs2 :P
(Score: 3, Informative) by frojack on Tuesday September 02 2014, @01:12AM
Actually it started well before Oracle got involved.
They paid the salary of some developers, true. If you want to toss out all of Oracle's contributions to Linux just because they are Oracle you would end up doing serious damage. The principal implementers came from Suse, had worked on ZFS, and ReiserFS (for suse), and went to work at Oracle and was given free reigns to develop the system.
Btrfs is a true open source project - not just in the license, but also in the community.
http://lwn.net/Articles/342892/ [lwn.net]
No, you are mistaken. I've always had this sig.
(Score: 1) by nishi.b on Tuesday September 02 2014, @02:37PM
Right with you on this.
I hope to use Btrfs and its snapshots, checksums and all.
Reading LP's projects to use hundreds of subvolumes means it will be such a mess that I won't know what I am using, which subvolumes I need to backup...
And he says ext4 will be supported... by putting btrfs filesystems in a loop file !
Please keep this on ONE distribution if this feature-set is right for you but do not push this everywhere...
(Score: 2) by nukkel on Tuesday September 02 2014, @08:44PM
btrfs is good, but ReiserFS is a real killer!!
(Score: 3, Interesting) by zocalo on Tuesday September 02 2014, @12:32AM
UNIX? They're not even circumcised! Savages!
(Score: 0) by Anonymous Coward on Tuesday September 02 2014, @08:05AM
Horse and carriage, love and marriage, lalalaaaaa...
(Score: 0) by Anonymous Coward on Tuesday September 02 2014, @12:40PM
Sorry, I can't hear you. Must be a Pulseaudio problem.
(Score: 3, Insightful) by jimshatt on Tuesday September 02 2014, @10:57AM
(Score: 2) by LoRdTAW on Tuesday September 02 2014, @05:17PM
Great post. You hit the nail dead smack on the head. BUT I do see the point of where this is leading to. It is both good and bad. I'll explain later.
To me the goal of RH with these ambitious "one ring to rule them all projects" is to make Linux distros nothing more than a skin on top of a common core OS. If you ask me they want Linux to be the next OSX. A Unix like kernel and functionality under the hood but with a common set of libraries and services on top all the way up to the UI. Gone will be X and other legacy API's/daemons and say hello to GNOME OS(Wayland, Pulseaudio, dbus, systemd, GTK, Gnome desktop).
Linux will no longer be a collection of software packages but rather a more complete core OS. Distros no longer have to worry about maintaining separate kernels, libraries, packages or repositories. Instead they pull in the Linux core OS, include whatever packages their version targets (education, Audio/music, Video, business/office, EDA, etc.) and throw in a custom logo, background, bootsplash and icons. This makes the life of the package maintainer easier as they can now provide a binary package that will work on any distro. This of course will have the effect of culling the distro herd as now who cares if its Debian, Gentoo, Arch, Suse or Redhat? It's all the same thing with a different background, packages and icons. My only hope is they distill the package install process to OSX, have a
THE GOOD:
If their vision works, it is NOTHING but good for the *Desktop* community. That will help move Linux closer to the year of the "Linux desktop" than any supposed "killer app" or Steam can dream of. Look I am going to be honest and say that as much as I love Linux there is still too much inconsistency and niggling problems which the average user does not want to EVER deal with. I can deal with it, most of you reading can deal with it. It's a challenge with a reward for us. To others it's a "fucking headache that needs to die in a fire." as my perpetual Linux newbie friend puts it. I am not saying I agree with this path, I don't. But I get it.
THE BAD:
But, there is a catch. And this is where it can turn sinister. If the concept of a distro is wiped out, many distros will die. Some will live on but the ones who stand to survive are those who provide commercial support. And who is in a better position to do that than RH? So the balance of power will shift to companies providing support for the new "One True Linux" and RH happens to be in the perfect position to do that. What do you think Valve, EA or Ubisoft will want to publish games on (besides the PS4 and XB1)? You guessed it, the "standard" unified LinuxOS. It will also open the floodgates to more proprietary software looking at moving to Linux as the platform will be viewed as homogeneous and easier to work with. No more worrying if your customer is running the latest Ubuntu and you only certified your CAD software to run on a Redhat version from 2 years ago. It "just works" like OSX and Windows.
THE UGLY:
And another major problem this is giving the finger to our open source, Unix brothers in arms, the BSD community and Linux purists. They now have to live with desktop software that won't work on their operating systems without heavy modification, compatibility layers or worse yet, adopting the LinuxOS userland and jamming a BSD kernel under it just to keep up. Sure strictly POSIX stuff will port like command line applications but this will break the desktop.
So it is a double edged sword. One one side you make Linux easier for the masses but on the other you concentrate power to a few players, if not one player.
You want my ¢2? They should have looked back to what the Unix greats were doing with Plan 9's 9p: make every bloody thing a file. That would solve a lot of issues relating to interprocess communication, messaging, video, audio, networking, etc. Instead of reinventing the wheel over and over they should have thought "Hey, how do we break these problems down to a common denominator?" That common denominator being how things communicate which is what all these goofy protocols and daemons are trying to solve. The sad part is the problem was solved 20+ years ago when the Unix developers saw the limits of how things were and are to this day, done.
My favorite Quote:
(Score: 2) by LoRdTAW on Tuesday September 02 2014, @05:53PM
Oops, posted instead of previewing.
Last sentence of second paragraph: "My only hope is they distill the package install process to OSX, have a" should have read:
My only hope is they make package management as simple as OSX. Seriously OSX does that one basic task right because it is as simple as can be. It feels more more like the Unix way. No scattering files all over creation, no idiotic registry entries, no goofy installers and uninstallers, or package managers. Installing a package is drag and drop. The application list in Finder is analogous to your bin or program files directory. The folder icon is pulled from inside the directory automatically and clicking on the folder launches the program. So in essence you just double click the programs directory to run it. Perfectly simple. A Linux package should be a tar.gz that in uncompressed to /user/apps/ or ~/apps/. Then you simply double click it to run. The GUI menu just needs to search those two directories to populate its program list. To categorize them you make a directory tree with links, the Unix way.
(Score: 2) by Marand on Tuesday September 02 2014, @09:24PM
The problem I see with the GNOME OS goal is that it's like trying to kill a mouse with a nuke: it might get the job done but you're destroying everything else in the process. Simplifying cross-distro installing for third-party packages is something that's been done before, multiple times, and never required destroying distros and handing the keys to the kingdom over to Redhat. 0install [0install.net] is one example that worked pretty well (especially coupled with rox-filer [sourceforge.net]) but didn't catch on; we've also got Docker [docker.com], which is the current golden child of this sort of movement; and even Steam shows that it can be done without obliterating the underlying system. Games have been packaging static-linked binaries to circumvent package management for years, and some even provide deb or rpm files. Then you've got things like Debian's checkinstall command, which lets you take a non-Debian installer and automagically make a .deb out of it for safe, clean removal later.
It's a great argument to make, citing ease of third party distribution and getting proprietary software companies on board, but it can (and has) been done without going the "GNOME OS" route, and should really remain that way.
The really sad part is, Linux was moving steadily toward that route for a long time, including things like adding the /sys/ mount, and then something happened. I don't know if we got an influx of a new breed of Linux user, or ex-Microsoft employees, or what, and they brought with them a desire to tear all that down and make everything binary logs, registries, blobs, hidden functionality, single-user systems, etc.