Gave Gnome another go
deleted it. Never sure what its meant to do - stop me getting near my computer?
As has become par for the course with Fedora, the latest – Fedora 21 – has arrived months behind schedule. To its credit, it's well worth the wait. This release marks the start of the Fedora.next project. The big change is that Fedora 21 is available in three flavours: cloud, server, and workstation. All three build on the …
Gnome was gradually driving me mad. I didn't delete it, but I did "apt-get install xfce4" and made that the default instead. I'm much happier now, though I have to say that Xfce on Debian didn't come with sensible default configurations: I had to fiddle around a bit to make it do something sensible when the laptop lid is closed, not run a ghastly 1980s-style "screensaver", allow the local user (possibly my 10-year-old daughter) to shut down without giving the root password, and a few other things. Nothing very difficult, but someone who is new to Linux would probably just try Gnome for a bit before deciding that Linux is crap and going back to Apple/BSD/Microsoft/whatever.
Ideally, distributions should actively maintain an alternative to Gnome and encourage people with low-spec hardware to use it instead of Gnome. Having alternatives in the package repository isn't good enough, particularly when there are dozens of them and no guidance on which one to install and how to configure it.
Anyone else wonder why the new binary logs in this systemd monstrosity require a qr code lib and an http server, just to deliver filtered log events? Even when bypasing journald and using syslog?
This post has been deleted by its author
systemd seems to attract some pretty shrill complaints but I fail to see the issue with it. The authors put out a copious rebuttal of the common complaints that makes a lot of sense and of course it's far more powerful than standard init scripts, just as reliable, and means users get a desktop faster and servers are ready sooner.
Here is a non-rant, detailed look at what is wrong with systemd:
"Here is a non-rant, detailed look at what is wrong with systemd:"
Most of the points are largely nitpicking about perceived logical inconsistencies in a selection of pro-systemd arguments. Not the same thing at all as saying what's wrong with the tool itself.
Nitpicking aside, answer my original questions about the necessity of a qr code lib and a http server for journald. Replacing text with binary, a binary that filters what you get see is bad enough. And the oft quoted "just replace it with syslog" is not a workable option, it merely sends the filtered events to it. What purpose does a qr code lib and a http server serve here? Take your time composing your answer, I'll wait...
Of course the broader question is why does so much of systemd, which is replacing the OS incrementaly, need to reside in user space? Why the single point of failure in PID1?
The article I mentioned destroys all the arguments currently being made as to the need for systemd. If it is not needed, why use it? Are you also a fan of PulseAudio, another Poettering epic fail?
"Nitpicking aside, answer my original questions about the necessity of a qr code lib and a http server for journald."
Simple: they aren't necessary. Fedora's systemd does not require an HTTP server:
[adamw@xps13 ~]$ cat /etc/fedora-release
Fedora release 21 (Twenty One)
[adamw@xps13 ~]$ rpm -q --requires systemd | grep http
Fedora's systemd is built against qrencode, but that's a Fedora choice:
[adamw@adam anaconda (master %)]$ cd ~/local/systemd/
[adamw@adam systemd (master %)]$ grep -R qrenc *
configure.ac:AC_ARG_ENABLE(qrencode, AS_HELP_STRING([--disable-qrencode], [disable qrencode support]))
Build systemd with --disable-qrencode if you don't want it to have a qrencode dependency (or in fact just don't build it with --enable-qrencode , disabled is the default). The benefit is some convenience when setting up keys for Forward Secure Sealing of journal contents; see the --setup-keys argument in 'man journalctl' for details.
As for the HTTP server - again it's a compile time option, --enable-microhttpd to turn it on. Fedora does in fact build with it, but then splits it out into a subpackage so the core systemd does not depend on microhttpd. The subpackage is systemd-journal-gateway . As that name suggests, the feature it supports is allowing remote access to the journal; the thinking was that there's a handy protocol lying around for reading information over a network which everything and its dog supports, so why not just use that instead of inventing some new protocol for remote accessing journal data? See http://www.freedesktop.org/software/systemd/man/systemd-journal-gatewayd.service.html . Of course, even when the package is installed this is not enabled by default, and the package is not installed by default.
"Of course the broader question is why does so much of systemd, which is replacing the OS incrementaly, need to reside in user space?"
Erm. Why would you want to build it into kernel space?
"Why the single point of failure in PID1?"
There's *always* been a single point of failure in PID 1. That's sort of what PID 1 *is*. Note that much of systemd does not run as part of PID 1. systemd-journald, for instance, is pid 559 on my system as I write this. systemd-udevd is pid 606. systemd-logind (the daemon GNOME uses, for the 'GNOME requires systemd!' haters) is 871.
The QR / htp code, is a disabled-by-default feature called forward secure sealing which is an anti-tampering protection that stops hackers from erasing your system logs or tampering with them without your knowledge. The QR is a convenience for people to quckly validate the server for tampering from a mobile device. The sort of thing an admin might find usefule from time to time. This is supposed to be bad how?
If you don't need it, don't enable it. Sheesh.
And running stuff in user land is a Good Thing. In fact, it's the whole reason for systemd's being. It's the bootstrap for userland.
And no the article you mentioned does not destroy the arguments for systemd. As I said, it's largely arguing against the arguments and says precious little about systemd itself. And ironically considering your link you've engaged in a little ad hominem yourself with your rant against Poettering and Pulseaudio.
> The article I mentioned destroys all the arguments currently being made as to the need for systemd
Not really, most of the arguments are strawmen "disproved" by weak or disingenuous arguments, for example Fallacy #5 which boils down to "you dont need the systemd rewrite of init to use $LINUX_SPECIFIC_FEATURES you can rewrite it yourself to use them!". #4.4 also comes down to "you can rewrite large chunks of init" as well. Not a highly persuasive argument.
Or Fallacy #2 "distributions aren't adopting systemd on its technical merits they're doing it because the Gnome develpoers adopted it on its technical merits" hmm.
#2.1 "Perhaps the most widely-deployed consumer GNU/Linux distribution by sales is ChromeOS, but ChromeOS uses Upstart" is nicely phrased to ignore Red Hat which is adopting Systemd
> Are you also a fan of PulseAudio, another Poettering epic fail?
You're getting dangerously close to completly invaldating Fallacy #11 all by yourself.
systemd seems to attract some pretty shrill complaints but I fail to see the issue with it.
Perhaps an analogy is in order. Suppose you bought a house, and your solicitor did the conveyancing. You would be pleased in your new home. Then suppose you had a baby, and, instead of a doctor or midwife turning up, that same solicitor phones up and insists on doing the pre-natal care and delivering the baby. He isn't qualified yet but says he has just started a medicine degree. Later on, your car goes in for a service, and there is no qualified mechanic, just that annoying solicitor again. Afterwards, you notice he has welded the bonnet shut so you can't see what is under there any more. You are very annoyed so go for a pint, who do you think is standing behind the bar...
People do different jobs for a reason. Computer programs do different jobs for a reason. Unix/Linux has been designed this way.
"Perhaps an analogy is in order. Suppose you bought a house, and your solicitor did the conveyancing. "
Perhaps a slightly better analogy is in order. Suppose you bought a house built a month ago yet you discovered it was still built to 1970s codes. Yeah it's a house with a roof but it's lacking by modern standards - lacking dead locks, double glazing, insulation, satellite points, smoke alarms, burglar alarm, wiring and safety compliance etc.
That would suck right? It would suck even more if the 1970s defence force leapt out with dubious arguments - Ah they might say, the builder who built this 1970s house knew what he was doing. Not like that OTHER LOT with their fancy double glazing and asbestos free cavity insulation. No, put up with your drafty substandard house and don't ever dare or hope to improve it in any way. No, don't even question why other houses manage to be better than this 1970s house.
A far better analogy. Though perhaps you think Red Hat are a bunch of cowboy builders?
No, Red Hat are not a bunch of cowboy builders. They are a bunch of naive builders allowing themselves to be fooled by Fedora team. I looked at the Gnome interface on RHEL 7 and I was appalled to see a lame Fedora clumsily bolted on top of a solid server. They could have come up with a GUI for a real server in the way (yes, I know) Microsoft does. Instead of sysadmin tools we get "... configure your accounts in order to store your music and photos in the cloud..." stuff. On a corporate server ? WTF?!
The RHEL default desktop isn't a server GUI. It's a desktop GUI. People run RHEL on desktops; we have major customers running multiple thousands of seats on it. That's what the RHEL desktop is for.
Probably most people running RHEL as a server OS don't run a graphical desktop directly on the system at all; those who do usually run something pretty geeky and minimal, not GNOME. That's not what GNOME is for.
RH is working on 'server GUIs', but more in the line of webapps than graphical desktops running directly on the server machines. See, for e.g., Cockpit - http://cockpit-project.org/ , included in F21 Server - and the FreeIPA web UI - demo site at https://ipa.demo1.freeipa.org/ipa/ui/ .
"Instead of sysadmin tools we get "... configure your accounts in order to store your music and photos in the cloud..." stuff. On a corporate server ? WTF?!"
Fedora isn't a corporate server, it's a test bed. Secondly, nobody would be administering a network or a server with default desktop apps so your point is ludicrous - they would do so from a terminal or perhaps a admin app such as cpanel or webmin. Thirdly, anyone who was deploying a desktop to their users in an enterprise would take the time to customize it to their requirements - default, apps, icons, themes etc. to expose. Fourthly, it's the simplicity of GNOME which would make it so appealing in this regard - the less things there are for a user to break, the less it will break. And fifthly, if someone were desperate for another desktop, they can just install one via yum.
So no, they're not "naive". They know a damned sight more about their business than you appear to know.
No my response shows that I've taken the effort to read the issues are with init and why it might be desirable to replace it with something more robust, faster and more capable to support the requirements of modern desktops or workstations.
It's not the first time this sort of thing has happened either. Wayland is another recent example of people pouring on their irrational hate on something because it's different.
But the great thing about Linux is you can choose not to use a dist if you don't like it because there are bound to be others that do things the way you want. Or you could roll your own. Or you could whine and downvote people who choose to think Linux isn't perfect and are supportive of attempts to improve the experience.
Having binary rather than text format logs and only being able to pass-through filtered (WTF?) information doesn't strike me as being desirable in any way shape or form and I could only surmise that it exists due to the will of a fucking idiot.
But the great thing about Linux is you can choose not to use a dist if you don't like it because there are bound to be others that do things the way you want.
When all major distros have gone down this route best of luck with your replacement options.
The filtering of events to the logs is the key problem with journald.
What is filtered out, and why? Sending the logs to syslog doesn't overcome the filtering.
Why filter? Who benefits from this?
I don't expect an answer from the systemd cabal other than "trust us", which I don't...
Actually, we're the guys who got the original Secure Boot implementation designs substantially changed to be less sucky. Without the RH representatives in those discussions, all systems with Windows 8.1 pre-loaded would likely be as locked down as an iPhone.
If you guys were that great in changing the design of Secure Boot why didn't you change it to allow replacing the Microsoft public key ? Let's not forget that by doing what you did you prevented the rest of the Linux community to stand up and fight. Yes, RedHat caved in too but hey, they are a business that must care to their shareholders. This was a real sell-out because sooner or later all systems will locked-down as an iPhone.
"If you guys were that great in changing the design of Secure Boot why didn't you change it to allow replacing the Microsoft public key ?"
The UEFI specification's Secure Boot definition does not make any requirements whatsoever regarding who should be the provider of the platform key, or any other key, or whether any keys at all should be provided in any particular firmware implementation.
The Microsoft labelling requirements explicitly require that the system owner be able to replace the Microsoft key. Look, they're right here: http://msdn.microsoft.com/en-us/library/windows/hardware/jj128256.aspx
"On non-ARM systems, the platform MUST implement the ability for a physically present user to select between two Secure Boot modes in firmware setup: "Custom" and "Standard". Custom Mode allows for more flexibility as specified in the following:
A.It shall be possible for a physically present user to use the Custom Mode firmware setup option to modify the contents of the Secure Boot signature databases and the PK. This may be implemented by simply providing the option to clear all Secure Boot databases (PK, KEK, db, dbx), which puts the system into setup mode.
B.If the user ends up deleting the PK then, upon exiting the Custom Mode firmware setup, the system is operating in Setup Mode with SecureBoot turned off."
"Let's not forget that by doing what you did you prevented the rest of the Linux community to stand up and fight"
What do you mean by 'doing what we did'? And how did that prevent anyone else who gave a crap from pulling their finger out and doing something useful?
While I had tested Fedora 21, I found myself re-installing Ubuntu 10.04 last night to have hardware support for my 20 year old HP LaserJet 5P with a usb (rather than the original parallel port) connection to my computers.
Some of us do not trash hardware that is still useful. Hence I need the older linux kernel that still provided that support.
Thanks for the review, glad you enjoyed F21 overall!
"Given the base system shared by these new flavors, I hoped there might be a plain, base flavor - something akin to Debian's Minimal install, which would let you build a more customized desktop. But so far that's not something Fedora is doing. In fact, the Fedora project emphasizes that the base set of packages is "not intended for use on its own"."
Where did that "not intended for use on its own" quote come from? It might need tweaking. You certainly can do a minimal install of Fedora, and it's fairly common. It won't be a Product, but it's a perfectly valid Fedora install. To do it, use the Server network install image, go to the Software Selection screen and pick the 'Minimal Install' option, it's all the way down the bottom of the left-hand column. (Note for the haters: Minimal still uses systemd, NetworkManager and firewalld. Sorry, haters.)
"With Fedora's installer it isn't immediately clear what you need to do – or even that you need to do something"
Well, er, there *is* a rather bright orange bar at the bottom with a little attention icon, that says "Please complete items marked with this icon before continuing to the next step." And the spokes you have to complete - only Installation Destination is required, usually - are marked with that attention icon. Was this not visible enough?
"GNOME on Wayland is still very rough around the edges and there are a number of apps that won't work with Wayland (some of which might never be ported to use the Wayland protocol)"
Note that most of those should still run in a Wayland *session*, via XWayland. In fact I think it's still the case that most apps run via XWayland in Fedora's Wayland GNOME session by default - GTK+ 3 does *have* native Wayland support now, but it was still so crashy the devs decided not to use it by default yet.
"Other Fedora 21 highlights include some SystemD updates"
Nitpick: it's systemd, all lower case. The 'SystemD' spelling tends to be associated with the comment thread monster raving loony faction, so the devs get sensitive about it.
Thanks for all the kind words!
The support lifecycle for Fedora is so short it's not worth bothering about unless you're just playing. There's nothing wrong with just playing of course, but for those of us who use our machines as a tool to get on with our actual work these short-lived distros are a waste of time and effort.
You could take that approach but, with a couple of minor exceptions which took at most a couple of hours to fix, I have been using Fedora since about FC3 or 4 and upgrading it along the way. It's good enough for what I do with it (everything at home) and I see no reason to change.
If you'd rather use something with a long life cycle, that's entirely up to you, but it's kind of presumptive to assume you speak for everyone "who use our machines as a tool to get on with our actual work".
Fedora's life cycle is ~13 months (technically it's 'until one month after the next release but one comes out' - F19 goes EOL in a couple of weeks). That's not like RHEL, or anything, but many people find it perfectly viable for getting all sorts of 'actual work' done, and upgrades generally work well these days.
"If Fedora sends you here to do PR work this means we got their attention."
Unlikely. Only those who agree with the devs receive attention. Everyone else is a raving loony. It's the old NSA conundrum: they demand unquestioning trust and faith, and anyone who doesn't deliver is the enemy. Of course, that very attitude is largely responsible for the increasing lack of both trust and faith...
No distro that sports syStEMD will get dime one from me. Those that offer a fully functional path forward without it will. There's kinda not much I can do except that, eh? The chances of getting the devs to sit down and come to any sort of reasonable, mediated compromise with the community are virtually nonexistant. It's their way or the hiway, and if you won't put your absolute trust and faith in them, then you're the enemy.
I'm oddly okay with that. After a decade of being a heathen traitor to the Windows admins because I ran heterogenous environments and championed RedHat/CentOS, I think I can handle being branded an enemy of the state by RedHat's elite as I go forth and seek a better alternative for the next decade.
<sneering opensource hipster nasal voice>
Well if you don't like it you can write your own versions!
<barely audible volume>
Just make sure you use the versions of udev etc before we made them unusable as self standing entities...
</barely audible volume>
Don't have any talks about your own versions or me, my RH mate and that guy that merges shitty code into the mainline kernel under Linus' nose will turn up and bully you! Oh why is everyone in opensource so nasty to me booh booh
</sneering opensource hipster nasal voice>
Er, if anyone was paying me to do PR, they'd want their money back, I think. No-one's sending me anywhere to do PR, my job is QA. This I do on my own time, God knows why. I've been posting here since it was a useful tech news site and not some sort of weird climate change sceptic backwater, and long before I worked for RH....
AdamWill inscribed for posterity thusly"
"I've been posting here since it was a useful tech news site and not some sort of weird climate change sceptic (sic) backwater, and long before I worked for RH...."
My, My, now anyone who doesn't drink the systemd koolaid is global warming denier?
Surely you can come up with a better ad hominem than that...
I know this is the Interwebs an all, but as a bystander still trying to work out whether systemd is the spawn of the devil, or the kind of work that Fabrice Bellard might come up with over a long weekend...
please can we play the ball and not the player (aka minimise the ad hominems)?
[Fabrice Bellard is God, like Clapton is God. Is everybody OK with that? Because if you're not, you too are the spawn of the devil]
Systemd is both the spawn of the devil (Poettering) and something bordering on a useful technology. Here's a quick rundown:
1) The systemv init sytem in Linux is awful. Truly awful. A replacement is needed.
2) Poettering decided he had the One True Plan to replace it, and set about doing so...he even had some truly great ideas. He also had some real stinkers.
3) Poettering refused to listen to anyone about what might be good or bad ideas, and proceeded to push forward with creating a massive blob of interdependent and interconnected binaries that together form something suspiciously close to a second kernel living in userland. The thing isn't done yet, but if Poettering accomplishes everything he claims he is setting out to do, systemd will functionally become a second kernel: something that almost all Linux systems hook into and rely upon to work.
4) This places Poettering - and thus Red Hat - in a position of central control over Linux that has traditionally only ever been occupied by Linux Torvalds, with his control over the kernel.
5) Poettering is an ass, has behaved poorly when interacting with other devs, and has earned the wrath of Torvalds. In return, he has complained that everyone are big meanies and - more critically - they'll just all have to learn to fucking cope, because he's going to do what he wants, and damned be the first that says "hold, enough".
6) Which brings us to today. Today we are in the "in between time". today you can (sort of) build a distro without systemd, though you need shims to do so, and you'll need even more shims tomorrow. Poettering/Red Hat have completely cut off the rest of the community and have simply resorted to calling everyone who disagrees with them "loonies" because you can theoretically roll a systemd-less distro today. Red Hat claims that anyone worried about it "infecting" Linux and becoming a userland kernel are nuts. Look at it today, it's not there yet! But don't talk about tomorrow, or what the developers have discussed, etc. Because that isn't in the code, so there's no need to worry your pretty little head. Trust Red Hat and Poeterring. They know best.
What the whole debate boils down to, at it's core, is that Poettering and Red Hat honestly believe they know what's best for us (which, completely incidentally, is something that places Red Hat in control of virtually the entire Linux ecosystem). Thus any objections are considered the objections of people mentally unfit to make objections.
There are lots of bent feelers on all sides and discussions about the technology itself - which has both good and bad elements to it - are completely subsumed by politics.
The systemd debate is, at it's core, about who is in charge of Linux. The community, gated by Torvalds as a last sanity check at the kernel layer, or Red Hat and Poettering? Is it okay for Linux to become a web of interdependent choke points and lock-in gated by someone who is outright hostile to the Linux community? Is it okay for Linux to almost entirely come under the control of employees from one commercial entity?
This is what the debate is really about, and I can even tell you how it will play out: LInux will fracture (again).
Linux itself is currently "Android" and "Linux". (Heretofore referred to as Android/Linx and GNU/Linux, as while GNU/Linux is not 100% accurate, it does reflect the idea that it's main goal was to provide something that was an open source alternative to Unix, and adhered to many/most of the Unix principles.)
Nobody really looks at Android/Linux and says "wow, that's Linux". So much has changed in the basic architecture, toolchains, even file layouts that it doesn't look, feel or behave like LInux anymore. It is its own thing. It is walking its own path.
GNU/Linux is in the process of a very messy, very public fracture into GNU/Linux and systemd/Linux. Systemd/Linux will be the dominant flavour, used by all commercial distros because that's where the bulk of the paid developers are going to be working.
GNU/Linux will splinter off and become a much smaller community (similar to Linux when it first started). They will have to completely scrap everything infected by systemd and rebuilt them. Many things (like Gnome) will take crazy amounts of effort to replace. It will be a decade, maybe more before a commercial distribution emerges based around GNU/Linux.
Once the first one is GNU/Linux distribution is launched, most commercial systemd/Linux distributions (those few that are left which aren't Red Hat) will switch to GNU/Linux. They will have by this point discovered that they don't have a voice in an ecosystem in which Red Hat has crowned itself emperor and will attempt to embrace GNU/Linux in order to seek differential and commercial advantage.
Eventually, one of those commercial GNU/Linux distributions will try to pull some systemd-like shit one more time and the cycle will repeat.
The systemd thing is an argument of philosophies. To whom does Linux belong? The company that puts the most money into development and strategised the most about controlling all aspects of it? Or to the community?
Is Linux a means to preserve freedom/liberté, or merely a means to a commercial end?
So there you have it: the Great Debate. I hope that clears some things up.
Well put Trevor, well put.
IMO, Poeterring and the systemd crew are the exact opposite of what I thought Linux to be when I started with it: Freedom.
I am so tired of the "roll your own" comeback they make. That will be hard when more apps are systemd dependent, like say the latest version of Gimp. It won't be long before apps sans systemd will be hard to find. They know this and couldn't give a shit about anyone's opinions/wants/needs but there own...
"IMO, Poeterring and the systemd crew are the exact opposite of what I thought Linux to be when I started with it: Freedom."
Well that's preposterous. Don't like systemd? Find a dist which doesn't use it. Or do you sort-of-like systemd but not all of it or the personalities? Fork it.
In summary, your freedom is intact.
Of course the reason systemd is being used by so many dists is because it is demonstrably better than either init or upstart. They didn't have a gun pointed at their heads.
"I am so tired of the "roll your own" comeback they make. That will be hard when more apps are systemd dependent, like say the latest version of Gimp. It won't be long before apps sans systemd will be hard to find. They know this and couldn't give a shit about anyone's opinions/wants/needs but there own..."
It's only as difficult as the willingness of people to get off their backsides and supply patches to do it a different way. If (for example) Debian chose not to use systemd, then Gimp and other applications would soon have #ifdefs which enabled some alternate behaviour.
That's always how Linux has worked and will continue to work.
Not everyone, a sizable number actually, would agree with your "demonstrably better" remark. Many sysadmins have stated their opinions all over the net about their dislike for systemd AND the lack of any listening by Poettering and crew.
And be clear, systemd is not an init replacement, it is consuming everything in sight, to the point of becoming a second kernal in userland. Logging, cron, bite by bite it is eating everything.
I do agree with your last point, Linux sans systemd will survive somehow. I am sending money to Devuan, and anyone else who will move forward without the systemd OS. Poettering did say that systemd 's eventual goal was to become an OS, did he not? Systemd is far more than an init replacement...
"Not everyone, a sizable number actually, "
And a sizable number are extremely favourable of it. Not surprising since it enables faster boots, lowers memory consumption, enables load on demand, increases security and a raft of other things.
Which may explain why Fedora, Arch, Gentoo, RHEL, OpenSUSE, Debian and Ubuntu all support or are in the process of supporting it.
"And be clear, systemd is not an init replacement, it is consuming everything in sight, to the point of becoming a second kernal in userland. Logging, cron, bite by bite it is eating everything."
So what? Those are all interrelated things. For example systemd implements timers so it can kick off services that do log rotation and similar but also allows them to run as a cgroup for better security. Doesn't stop you using cron but if you were security conscious you might like the added security of giving the least privilege to a job to do what it needs to do.
I still haven't read a coherent reason why systemd is bad. Most of them turn into ad hominem attacks against the main author, appeals to tradition or fear of change. But as I said, if you don't like it, use a dist which does it the old way.
It's not the first time this sort of thing has happened in the world of Linux.
"And a sizable number are extremely favourable of it. "
So the sizable number of people who agree with us should kowtow to the sizable number of people who agree with you because...well, because what, exactly? You have a bigger stick?
"Which may explain why Fedora, Arch, Gentoo, RHEL, OpenSUSE, Debian and Ubuntu all support or are in the process of supporting it."
No, they're in the process of supporting it because nobody in the Linux world has the financial resources to fight Red Hat on this. Nobody. Red Hat has played this very well; so many components are already dependent on systemd that maintaining shims or creating alternatives is a massively financially burdensome project to undertake.
Worse, the various teams under Red Hat's control seem to be planning, one by one to "integrate" with some level of dependency upon systemd. So the number of projects that have to be worked around, shimed or outright replaced is going up.
If you want a GUI on your Linux distro at all, you're pretty much fucked already. And it won't be long before those looking for a server OS are going to have to clean fork GNU/Linux away from systemd/Linux in order to keep free of it.
Any distro that wants to be commercially viable does what Red Hat says and hopes Red Hat doesn't turn their eye upon them and try to put them out of business. And that's the end of it today; Red Hat controls too much. Nobody else has any real leverage.
Linux functioned because no one group ever accumulated too much power, so relying on eachother's good will was backed by some plainly pragmatic reasons why everyone had to cooperate. That's gone now, and with it both the presumption and any pretense of goodwill.
Commercial interests are all that matter int he systemd/Linux ecosystem.
"So what? Those are all interrelated things."
Do one thing and do it well. The reason being not only to avoid gigantic monolithic entities with 2M lines of code, but so that no one person or group gains lock-in class control of the ecosystem.
"It's not the first time this sort of thing has happened in the world of Linux."
Yes, it is. Never before has an entity other than the kernel team held so much power that it amounts to a minimum of a veto, more realistically the final word on what will happen within the ecosystem. Through systemd's growing scope and lock-in, Red Hat holds that power They control Poettering and he controls systemd. Red Hat also control a number of other critical projects, many - most- of which are now or have announced plans to become reliant upon systemd.
Nobody has had this kind of power over the Linux community before. Not even the X team.
"if you don't like it, use a dist which does it the old way."
Which is what people are saying they will do, and advising others to do. So why get your panties in a bunch about it? People are allowed some semblance of choice and/or freedom, but only so long as they don't exercise it in public or attempt to make their case to others?
Man, I hope noone in your family every comes out of the closet. Your inability to accept that other people may value things that you don't - and may not value things that you do - would probably make that a pretty brutal affair.
Well that's preposterous. Don't like systemd? Find a dist which doesn't use it. Or do you sort-of-like systemd but not all of it or the personalities? Fork it.
#WorstStrawmanEver. Virtually any viably usable non-niche well supported distro has gone down the systemd rabbit-hole, how do you propose people find another?
"#WorstStrawmanEver. Virtually any viably usable non-niche well supported distro has gone down the systemd rabbit-hole, how do you propose people find another?"
Er, no it's not a straw man. It's the way things always happen in open source. Forks happen all the time - egcs, xemacs, x.org, libreoffice, mariadb, jenkins, mate, io.js. The list goes on and on. Some succeed and become defacto and some don't.
Sobbing over systemd is pathetic quite frankly. If it offends you so much, fork it or the dist or support someone who has.
Ahhh, so people should give up using Debian, Mint, Ubuntu etc and use MyShittyDistroIForkedAndCanOnlySupportOnWeekendsWhenImNotBusyOr
Practical, real practical. It's like every car manufacturer only making diesels and your answer being "if you don't like it then create your own petrol engine".
The statement people are making (that you just do not seem to get) is that the whole point of Linux is that, once you're outside of the kernel, you are not effectively dictated to as to what you should use as bits can be swapped in and out. Use package X for this, package Y for that. If these arseholes gradually make every important component rely on this shite called systemd then they have broken the fundamental tenet of system design whereby you do not tightly bind as it creates a shitload of issues. As someone else pointed out - who has the time and money to go round backing out all the shit this multi-billion dollar company has effectively forced on a community. Just because they've played their hand well it doesn't stop them being wankers.
I still cannot get over filtered binary logging. Mind-numbingly stupid.
"So there you have it: the Great Debate. I hope that clears some things up."
Certainly does, thank you. And no messengers or players were harmed (well, not much) in the production of those words.
Would it be unfair to say this is an attempt by RH (and/or one individual) to do for GNU/Linux Classic what Google did for Linux/Android with Play Services?
Or would that be a complete misunderstanding?
"Would it be unfair to say this is an attempt by RH (and/or one individual) to do for GNU/Linux Classic what Google did for Linux/Android with Play Services?
Or would that be a complete misunderstanding?"
I think it would constitute a misunderstanding. Think of it more as "Red Hat trying to to for Linux what Apple did for BSD." They want to own the product. A vestigial community can exist outside of their control - they don't particularly care - but all main components of a commercially viable Linux platform must come under their direct control.
If you want the equivalent of the Google Play store for Linux, take a look at Docker.
I'd say that's a complete misunderstanding. I can see where the idea comes from if you don't know anything about how RH works, but it looks nothing *at all* like that from where I'm sitting. systemd isn't really a Red Hat project at all. It's a Lennart project. Lennart isn't an RH plot; he's an engineer. RH hired him based on his Avahi/PulseAudio work. No-one ever told him 'hey, Lennart, Red Hat wants to own init now. Go write an init daemon, and make it a big one.' He just got interested in the area and decided to write something; same as he did with all his other projects. He'd have done the same if he'd been working for Wal-Mart.
systemd got adopted in Fedora because Lennart proposed it and fought the technical merits until he won. It got adopted in RHEL for the same reason. It came from the bottom up, not the top down - there wasn't some PHB somewhere coming up with a strategic vision for the future that involved an init system, it was an engineer pushing something he thought was good. If you want to look at the stuff that actually *does* come from Red Hat's strategic vision and gets dropped into Fedora, look at stuff like Cockpit and FreeIPA and to an extent the Cloud stuff (though some of that is engineer-driven, too). That's the stuff RH's 'strategic vision' is resulting in. Not init systems.
It doesn't really *benefit* RH to 'own' init (not that we do; even if you accept that Jude person's point that there are ~10 core contributors to systemd, half of those don't work for RH). That's not much of a selling point to any of our customers. It's just more work we have to pay someone to do. Have you looked at RH marketing lately? Have you seen how much of it is about systemd? That would be 'none'. It's all hybrid cloud this and container that and devops the other. I mean, I just went to https://www.redhat.com/en and clicked around the marketing crap aimlessly for about 10 minutes and the word 'systemd' didn't appear once.
Oooh! I finally found it, on page 6 of https://www.redhat.com/en/files/resources/en-rhel-whats-new-in-rhel-712030417.pdf , which is six clicks from the front page, one unobtrusive link on https://www.redhat.com/en/technologies/linux-platforms/enterprise-linux . Prominent, this ain't.
"It doesn't really *benefit* RH to 'own' init (not that we do; even if you accept that Jude person's point that there are ~10 core contributors to systemd, half of those don't work for RH). That's not much of a selling point to any of our customers".
First off, systemd is more than init, even now...and there are publicly stated aims to take it a hell of a lot farther yet.
Secondly, Red Hat "owns" systemd in that they employ Poettering and Poettering is the individual driving that project. Anyone who gets in his way will be smacked down with extreme prejudice.
Thirdly, it absolutely does benefit Red Hat to make as much as possible within the Linux ecosystem completely dependent on something they control, as it gives them the ability to wield far more influence within the community than they could simply through the developers they employ. Control the frameworks and you control the ecosystem.
Lastly, you make the rediculous assumption that Red Hat only acts in the interests of it's customers. Just becuase something isn't what Red Hat's customers want doesn't mean it isn't in Red Hat's interests. Look to Microsoft for some great lessons in this: Microsoft regularly acts against the interests of it's customers, developers, partners and staff because it gives Microsoft leverage, the ability to create lock-in and/or dependance or allows them to move people away from less profitable models to more profitable ones.
Red Hat aren't the good guys any more.
More's the pity, as I really am not all that good at any of the other distributions. Walking away from Red Hat is meaning learning the nuances of Slackware (and possibly this new Debian fork) instead of relying on decades of Red Hat knowledge that has become so deeply integrated it's autonomic.
But there's more that matters in this world than commercial interests and I still care about more than maximizing my own revenue. Pity that's not a sentiment much shared any more.
"Have you seen how much of it is about systemd? That would be 'none'. It's all hybrid cloud this and container that and devops the other. I mean, I just went to https://www.redhat.com/en and clicked around the marketing crap aimlessly for about 10 minutes and the word 'systemd' didn't appear once.
Oooh! I finally found it, on page 6 of https://www.redhat.com/en/files/resources/en-rhel-whats-new-in-rhel-712030417.pdf , which is six clicks from the front page, one unobtrusive link on https://www.redhat.com/en/technologies/linux-platforms/enterprise-linux . Prominent, this ain't."
Explain to me why Red Hat would spend much time advertising something that is not only controversial but a clear attempt to gain veto-class lock-in powers over the Linux ecosystem?
Even Microsoft isn't stupid enough to advertise the moves they make which fall into that category. That's the sort of thing you bury the lede on and then send out your "social media engagement team" to convince everyone that disagreement with the party line is nothing more than conspiracy theory. Three to five years later, you're welded into the pole position and you can't be removed.
Er, no. Believe it or not, not everything is about systemd. I was talking entirely literally about climate change. I stopped reading this site regularly back a few years ago when they'd post a new anti-climate-change wibble from Lewis Page every day - he's still at it, see http://www.theregister.co.uk/Author/93/ , though he seems to have slowed down a bit.
Lewis is one guy. (Though A.O. does post a lot of climate denier bullshit too.)
Simon Sharwood and Richard Chirgwin know their science and post proper articles with reasoned analyses. Until recently Rik Myslewski did too. Believe me, climate change denial is not institutionalized at The Register, any more than Worstall's Randianism is.
Yes, I would rather use something with a decent life cycle and so I don't use Fedora (and one year isn't just not long, it's ridiculously short.)
If your "actual work" is developing Fedora, then I can entirely see the point - if not, why bother? For the record I generally use CentOS on servers purely because of the support lifecycle and have used Gentoo on my own desktop for over a decade. This way I can keep an up to date system through incremental changes as it suits me without having the hassle of blowing everything away each year and starting again.
Fedora Core is a development bed for Red Hat Enterprise Linux. If you want long term support, use that. It means foregoing cutting edge stuff but on the flipside it's more stable & predictable as a result. Or Ubuntu LTS. Or Debian stable. Or a bunch of dists which aren't moving so much.
Fedora isn't a production distro, it is the community based test/release distro for Red Hat. It is deliberately packed the most up-to-date (and least stable) features and software. Features that are successful in Fedora 21 might get into Red Hat in a couple of years. Nobody is running production landscapes on Fedora. This is all explained better over at the Fedora project itself.
@Jim 59; I'm aware of that, I've been using Linux since long before Fedora existed. My original point still stands and you've really just agreed with it; whilst Fedora is fine for people playing with worse ways of reinventing the wheel, it's hardly a suitable distro for anyone actually needing a stable, reliable operating system on which to carry out non-linux-distro-developing actual work.
I love Fedora21. Anaconda is as simple as can be. It is idiot proof. and if you can't press onto some icons in front of you, then what can we say. My granddaughter, age 11, did my installation. She never saw Linux before, and only uses FaceBook, Instagram and Twitter.
There are buttons that must be visited if they have an orange flag with the icon. How difficult is that, to click, respond by reading two or three lines in very simple English.
Yes, my granddaughter did the Gnome installation. And I am responding with the version she installed.
We all know Fedora is intended for the Facebook people.
Fedora is intended for advanced Unix users, those able to do troubleshooting and bug reporting. Mint is intended for the regular user. Slack is intended for those who flagellate themselves with pine twigs in a steam room while listening to Mahler. Ubuntu isn't actually intended for users at all.
I had two tries at F21+XFCE+PAE....both were MUCH more work than using the "Full Install" DVD for F20. (Target machines are AMD/Radeon/32-bit)
1. Use XFCE Live and install to hard drive. No software selection at install time (that I saw). So lots of work afterwards with "yum install etc etc etc")
2. Use Workstation install. Very limited software selection at install time. Ugly Gnome 3 desktop. So in addition to lots of work afterwards with yum, I had to install XFCE.
Neither install noticed the 6GB of main memory, so in both cases I had to install the PAE kernel manually.
Install number 2 was fine in the end, and I'll be keeping F21.......but going back to the full 5GB DVD with almost everything on it (like F20), would be welcome here.
First they are the lone idiots in the forest sticking with
Metro GNOME as the default. Now with 21 they've dumbed down the install to where you have to choose between server/workstation at download time, and forced live install on everyone unless you use server - but then you have a bunch of packages missing if you want a GUI.
All this talk about improving the install experience but it keeps getting worse for those of us with a clue. It isn't as though they can't put a big "EASY INSTALL" button and keep improving that experience, without taking away the features for those of us who know what we're doing and what we want.
I would be totally unsurprised if a few releases down the road they eliminate the ability to change disk layout entirely. It'll just look for a Windows install, shrink the partition, create a new one in the open space, and install there. Because it'll be easier for the kind of drooling morons who think GNOME 3 is clever they're apparently targeting.
You can install any package set from the 'Server' network install image, it is really a generic network install image. The Product stuff is a first cut in F21, it's still being shaken out; this kind of quirk will be straightened out for F22/F23.
Red Hat lost me at version 7. It had such a convoluted install I had to redo it three times to make sure I had what I wanted. So why waste time on 21. I might have tried it, but it uses systemd. Those who like systemd should do a forensic exam of a failed system, do this using a linux that does not support systemd. Remember ALL systems will fail at some point. So go ahead do a recover on systemd.
I await the tale.
Gnome was made by The trash icon hater Nazis that want to rectally forced feed people (CIA style) with their different way of doing things. Terrible. And they even dare to make it evolve with time to make it better. Teeeerrrrriblllee.
So YEARS(!!?!) after the creation of Gnome 3 we still have comment about how bad Gnome is, how people are stupid using it and how people will have to walk on their dead body to make them use this unholy GUI.
Use something else. Quit trolling.
But honestly we see less of them.
Now we have a Greater Nazi Scare : SYSTEMD. Each article on SysTeMD creates a flood of FUD, trolling and half disguised personal insults to the developers (or Developer I would say, if you see who I mean...).
That's pitiful (and don't get me started on the "insults" part, that's such a shame)
If you are really so angry about SysTemD just take your finger out of your a$$ and start DOING things with the people developing alternatives. Stop trolling. DO something.
In all of the anti-systemd ranting and raving that I've seen here and elsewhere, there is always one thing missing: no mention all at all of any alternative means to actually provide the services of systemd. Why is that? I want those services, and others do too.
I've been using Linux in production for ages (>17 years). I've used it to do countless different things that I needed to get done. I was highly skeptical of systemd when it came out and was quite irritated using it in the beginning. However, I did take it upon myself to actually read the documentation for it and use it so that I could give it a fair evaluation instead reading crap like this thread. My verdict: It's not perfect, but it's better than what came before it. It makes my work *easier* and provides benefits that the older init system simply could not. If you don't agree, that's fine. Everyone's work is different.
Criminy, the paranoia (RedHat) and anti-Lennart diatribes here are simply ridiculous. Lennart's an ass. Yep, I think so too. He can be his own worst enemy. However, he's a competent ass and that's more important to me.
I've never understood the hate for Pulse Audio. Never had an issue with it. It solved perfectly the issue of providing audio to numerous x-terminals. The solutions up to that point simply sucked.