Losing graciously | Hacker News Reader

Losing graciously

markshuttleworth.com

513 pointsash202 comments

Some individuals in the Linux community regularly criticize Canonical for "not playing along well with others" and taking Ubuntu in directions that go against "the community's wishes." I think those individuals are misguided. Canonical does try to collaborate with other projects and distributions. It's just that such collaboration is not the company's priority.

Canonical's priority, as best as I can judge, is to make Ubuntu the world's leading Linux distribution for desktops and other platforms -- cloud, tablet, phone, etc. So they will use whatever F/LOSS code works best to achieve this goal, regardless of whether it's internally or externally developed. They are not letting 'ego' and 'pride' get in the way of achieving their goal.

If using systemd will help them achieve their goal more than using upstart, they will use systemd. Ditto for Unity versus Gnome, and for Mir versus Wayland. For them, it's not about "winning the argument," but about "winning the #1 spot."

Unity and Mir have been two good choices.

Mir isn't really rolled out enough to settle the score on that one.

Totally agreed on Unity though - it is a breath of fresh air compared to most alternatives on linux.

Unity was rolled out too early, and in a hardly stable state. It sacrifices desktop usability for consistency across devices: desktop, tablet and phone. I disagree that desktop computing design should be constrained by the limitations of touch-based devices. So, rather than a breath of fresh air, I see it as a disruptive change in UX, intended to win in the long term, but at the expense of short-term usability. Ditto for the move to Mir.

But I don't resent Canonical for this -- I use Linux because I can choose the experience (KDE 4 at the moment). I just hope the short-term sacrifices in terms of usability and fragmentation don't harm adoption too much.

I was on that same wave length for a long time. But I gave it a year then rounded up the usual suspects (xfce, gnome, kde) and was surprised to see how much I missed unity.

It sacrifices desktop usability for consistency across devices

I don't know what kind of usability impediment you're talking about? To be honest, I'm kinda tired of hearing this, because it seems to come from people who never really used Unity. I have extensively used KDE (v 2), Gnome (v 2), xfce4, Unity and now i3. I tried Unity simply because I was puzzled by the conflicting arguments. Power users complaining so much, while newbies were raving. I thought I would test drive it for a month just to see, I ended up using it for a year. Unity isn't perfect (no graphic environment is), but it does have its highlights. Most of the essential functions are documented in the cheatsheet that appears when someone keeps the WinKey pressed.

In my opinion, most power users who complain never took the 2 minutes required to understand how that desktop works and presumed that all they had to do was to plug in their mouse and click away. If they'd taken the time, there would be less complaints, since Unity is rather keyboard centric. Newbies on the other hand thought "let's learn Linux", read the manual, and came out ecstatic.

Just an anecdote of course, but on my not very beefy laptop Unity is the only environment I've tried that runs slower than Windows. Using it it seemed great and I'd like to use it, but unusably slow for me.

I don't know what kind of usability impediment you're talking about?

Not being able to use the programs installed on my computer?

That seems like a pretty big usability impediment to me. I have a Unity and non-Unity laptop. The problem that Unity allows less than ten apps on the side. It's true that I can search for the things - but that requires me to know the name of the many oddly named Linux apps out there. And yes, I've spent more than minutes, in fact I've spent days using and attempting to customize the thing.

The screen mechanics are ghastly also imho but I could learn them if, like, the GUI let me actually run my installed application.

Sure it works as a kiosk or those who just start a shell and leave. But if you're really using seriously, why pontificate about it.

It's true that I can search for the things

That is the primary method of using it, it's not the alternative.

The problem that Unity allows less than ten apps on the side

For one you don't need to know the name of the program, you just need to know a few keywords about what it does. e.g. type in "brow..." in the dash and see it proposing Firefox and other browsers you may have installed. In case you still have no idea what your program does, nor what its name is, you can still navigate to find your programs the old fashion way, using the dash and filters. I would assume that after using a program once or twice a user can either drag its icon to the side or simply remember its name. And if they do neither of those, the dash keeps a history of most frequently accessed programs. While using Unity I had 3 icons on the launcher and I _never_ even used them, and I run many apps at once.

Not being able to use the programs installed on my computer?

This is the first I've heard of Unity conflicting with a program and that might be a bug in one or the other.

But if you're really using seriously, why pontificate about it.

Since many are pontificating against it, I think people deserve to be exposed to counter arguments. A lot of people have taken upon themselves to decide that Unity is this and that, without even using it properly. They spread fud as truth, instead of disclaiming it as their own opinion. And frankly when you see the arguments, it is obvious that they haven't used the thing for more than a few minutes. It just strikes me as odd that people would keep doing this when there is ample evidence that the GUI is positively received by most people who give it a chance. If you don't believe me, ask System76 or Zareason, after all their business is directly tied to having people satisfied after they buy a linux box.

I view your sentiment as a little shortsighted. On a typical desktop you end up with over a hundred of applications installed and you end up using only 10 apps daily. A typical Windows Start menu gets extremely cluttered and extremely hard to search for the app you want. That's why keyboard-enabled search works much better.

OS X does the same thing. It has a launcher on which you pin the most used apps and for everything else you need to search it with Spotlight, or go to the Applications directory in Finder. I haven't seen many people complain about the usability of OS X in this regard. IMHO, the Start menu is a broken abstraction.

As for the UI and that bullshit on being optimized for tables ... I actually like that Unity is so space efficient. For example, why in the world would I need a window top bar, if the window is maximized? Vertical space on 16:9 screens is precious, Unity optimizes for it and I find that to be awesome ... like, one of the reasons many people like Chrome's UI over Firefox is exactly because of this, yet Unity brings this design decision for all apps.

Yes, almost same opinion. I've got feeling about unity people are really trying to step forward linux desktop UI. Not just copy and paste then edit something.

- remove cascade style start menu. - quick search interface for application, files. - well organized keyboard shortcuts and inline shortcut help - make free vertical space as possible as - it's simple. users only need to understand windows key, left launcher.

Yes, It was buggy and slow at the first time. Maybe Canonical had pushed so much earlier. But it's now working well and really usable.

KDE 4 has all the benefits of Unity as well as the ability to be customized. For users that like a 'start' menu, you can have it, but you're right that keyboard launchers are just better most of the time. KDE has two built-in, and every other DE works with Synapse, which is much faster than Unity for keyboard-based launching.

The UI is optimized for tablets and phones, but not because of space efficiency. In fact, Unity is less space efficient, because it demands that the launcher be so huge (to afford touch-based interaction). In every DE I've ever used, I actually hide the top/bottom panel, but that's not possible in Unity (the top panel affords no customization whatsoever, AFAICT). So, in practice, Unity is consistently less space-efficient with my vertical space than every other DE, in addition to being less customizable.

> I don't know what kind of usability impediment you're talking about?

Having a single menu for all apps at the top of my 1920x1080 screen was a usability issue to me.

Since folks don't seem to know what I was referring to, I wrote up my experiences with Unity. The tl;dr is in the paragraph second to last ('But Why?'), for folks that don't want the nitty-gritty. http://killring.org/2014/02/17/unity-shortcomings/

I rather liked Unity but bug #739184 means I can't use it. Hopefully, someone sorts that out by the LTS 14.04 release.

What's so great about Unity? I think the latest version of Gnome is much better.

For small screens, it preserves a lot of real estate. When most notebooks are limited to 1366x768 screens, it's an excellent choice.

To emphasize, it preserves a lot of vertical real estate.

Yes, really.

Thanks for providing some insight. I guess that's not a concern for me, since my laptop is 1920x1200.

Fair enough, but most people are using Ubuntu on a desktop or laptop with moderately large screen (or is 1366x768 on a laptop considered small?)

Can you actually run Ubuntu on a phone yet?

I haven't tried Mir yet.

Unity was a terrible choice. The early versions crashed a lot. Every time I've tried to use it, I get frustrated and end up removing it within a day. There's some magic key combination you have to know in order to do something as fundamental as opening multiple instances of an application.

The "start" menu equivalent in more traditional desktops has a list of installed applications by category. Browsing this to figure out what applications you have is an excellent way to familiarize yourself with your system. The "search" function in Unity requires you to know what you're looking for; there's no way to discover what's installed by browsing.

Unity would be fine as an extra option. It might even be a good default for smartphones. But I jumped ship for Linux Mint when Ubuntu tried to shove Unity down everyone's throats by making it the default.

For the first two releases where Unity was the default, I tried to like it but had to give up within a week because it was simply too buggy, given the Compiz plugins I wished to run. (Had I used it stock, doubtless it would have worked better.)

In the third release featuring Unity, also the release where they removed the classic GNOME shell, I tried Unity again and was satisfied. It was better than the GNOME shell I had been using theretofore.

I do not believe Unity to be perfect, but I do enjoy using it more than almost any window manager. (For completely unrelated reasons, I switched to i3 a couple of months ago, but of window managers directed at normal users, Unity is certainly my favourite.)

Unity was buggy in the beginning, but it has been usable ever since 12.04 LTS. I'm typing this on a MacBook that I use for work and when I'm in OS X, I miss Unity.

The screen is uncluttered, the shortcuts are great, the dashboard is great, the various special effects are finally not nauseating, the only thing bothering me is that I can't move that launcher on the right side, but lack of customization is a problem with Gnome in general.

Red Hat also wants the #1 spot. This conflict is a source of a lot of FUD and technical conflict

Red Hat wants enterprise/server. Canonical seems to be shooting for consumer.

The many thousands of ubuntu-server installs are probably eating into RedHat's earnings.

Maybe... but the companies buying support contracts and sending real dollars RedHat's way are going to be mostly the big shops, the enterprise shops, and I've never seen ubuntu-server in an enterprise shop.

What's an enterprise shop? I've worked at a big public company that used ubuntu-server.

One that values support from companies like Oracle, is a good first approximation.

If you define an enterprise shop to mean a company that doesn't use software like ubuntu-server, congratulations, you are right, but you have also made a completely meaningless observation.

Well good thing that I didn't define it tautologically like that then.

Instead, I suggested - as did your parent comment - that support contracts are very important to "enterprise" shops. That can be considered a defining characteristic in so much as nebulous terms have defining characteristics. And if you don't understand the importance of those support contracts, a lot of their behaviour will appear strange.

It's simply a fact that a lot of these companies officially support RedHat, often Suse, sometimes CentOS, but rarely so far for Ubuntu Server. That may change, but Ubuntu is definitely not there yet.

Ok sure, that makes sense, if the qualification is SLAs and the like. I bet RedHat dominates in that case.

Support isn't just about SLA's, its about your vendor actually providing software/drivers/etc for your environment. I think you will find many commercial applications only support recent versions of RHEL, SLES or in some industries Oracle Linux.

These products don't ship with source code, and they tend to fail spectacularly when put on machines that don't happen to have the "right" kernel, library and system tools. Frankly, having shipped binary products on linux in the past, it can be a real PITA just maintaining an application over three or four versions of RHEL or SLES (which have a lot more in common with each other than they do with debian based distros).

Two random links to prove my point:

http://www.3ds.com/support/certified-hardware/simulia-system...

Random hardware link:

http://driverdownloads.qlogic.com/QLogicDriverDownloads_UI/D...

I don't have a formal definition, but in my head "enterprise" is somewhere around the "10,000+ headcount" line.

I of course have only worked for a few enterprise shops; my evidence is purely anecdotal, and I know that.

An important part of the equation though as to how much this matters to Canonical or RedHat; was this shop you worked in paying for a support contract?

(Support contracts are perhaps not a defining feature of "enterprise", but they are a common theme)

Enterprise is where the people who choose the software don't use the software. Software is sold there not bought.

US Army (2007): “When we rolled into Baghdad, we did it using open source. It may come as a surprise to many of you, but the U.S. Army is ‘the’ single largest install base for Red Hat Linux. I'm their largest customer.”[1]

[1]: http://web.archive.org/web/20110521185400/http://www.linux.c...

Isn't "General Justice" a great name for a superhero?

U.S. Navy uses Red Hat too, though perhaps not to that same extent as the Army.

Baghdad is cool, I guess, I'd rather go to the space station with Debian:

http://phys.org/news/2013-05-international-space-station-lap...

That's because the US Army is still fighting the cold war, and thought it was "Red Hate Linux".

Probably not much. If ubuntu didn't exist, most of those installations would just be some other free distribution.

Depends... do they pay Canonical for support?

If yes, then, yes, they're eating into RH earnings

Otherwise, it's a battle between Ubuntu Server and CentOS

Canonical made around $50m revenue last year (and lost $20m). RedHat made $1.5 billion and was profitable.

It's a pretty big difference!

That does not address the above comment. Just because they aren't getting money for the installs does not mean that ubuntu isn't being used as a replacement for RHEL.

It addresses the 'buying support contracts' part.

CentOS and Scientific Linux are used as replacements for RHEL.

I think your statement is more true of CentOS than Ubuntu. Many Ubuntu server installations choose it because it is free and reputedly easy, and pass up CentOS because it doesn't have a lot of selling points that appeal directly to non-Linux users, nor RHEL because it costs money.

Also your comment reminds me of an argument against stealing music on the Internet, lol.

Canonical doesn't really compete in the traditional enterprise space, we're a cloud company, so you'll see us on AWS, HP Cloud, etc on the public cloud side and OpenStack for private clouds.

Although no doubt there are Ubuntu servers in the cloud like that, this summary seems unrealistic:

> we're a cloud company

Then what's up with Unity and the push towards a common interface? Cloud servers are not phones/tablets.

And BTW, I think Canonical (and Microsoft) pushes towards common interfaces have been huge steps backward, no matter how well intentioned, so given that, I wish Cananical were just "a cloud company".

I meant in the context of servers we're a cloud company vs. traditional servers.

What does that mean, though? That you're not really able to support the variety of hardware you get in real life?

Just spit balling here but, in general enterprise OS selection is driven by software requirements. And the requirements, for enterprise linux software, tend to be RHEL or Oracle Linux... i.e., not Ubuntu Server.

That would be my guess.

After which there comes a natural acquisition.

I'm one of those individuals. Color me impressed (with Canonical). They will all ways have this "correct" decision.

Their decision to deprive me of choice with GNOME2 sucks.

Isn't that the GNOME Team's decision?

You might want to try out Xubuntu. XFCE is even more lightweight that GNOME2 and I have really come to appreciate it.

XFCE is great and all, but isn't MATE an actual fork of GNOME 2?

Or just install XFCE. I'm continually confused when people recommend different operating systems to change the window manager.

Xubuntu is just a set of packages for Ubuntu. I use Ubuntu LTS and one of the first things I do on a new install is sudo apt-get install xubuntu-desktop .

Now, if you're on RHEL then Xubuntu would be a bit of a leap, but just installing xfce on Ubuntu is going to be a lot more painful than installing xubuntu-desktop which has a little more functionality out of the box.

You ubuntu users crack me up. How hard is this:

    # apt-get --install-suggests install xcfe4

Package info: http://packages.ubuntu.com/saucy/xfce4

Ubuntu's target audience are actually who think that command harder. Seriously.

You think that the people who have an opinion on XFCE4 versus Gnome are going to have trouble with one command?

Xubuntu is significantly different from stock XFCE

What is significantly different? I looked on the xubuntu site and I could not find much information. I did not realize that xubuntu was a lot different than normal ubuntu with xfce4 installed.

RHEL is primarily a server, not a desktop. All the desktops are in the repos already though, and Fedora has offical spins if you don't want to install them yourself for some reason. Gnome, MATE, KDE, XFCE, Cinnamon, it just takes one command to group install.

It has always seemed strange to me too. I have just always chalked it up to being an ubuntu thing.[1] I don't know if it has caught on in other distro-ecosystems but it seemed like it started with Ubuntu; which was peculiar because you do not have to deal with dependency hell--you have apt right there in that little black box!

The other ubuntu curiosity is installing `command-not-found` by default? I never thought you could have something worse than /usr/games/sl in your path, but then ubuntu introduced me to `command-not-found`.

[1] I say this as a 15+ year Debian user.

What's bad about "command-not-found"? Surely, it is more useful than sl.

It takes a couple of seconds to process if it's not cached.

If I mistype something, it takes ages for command-not-found to search the entire Internet for “emacss” or whatever, during which I am slowly driven mad by the system punishing me and forcing me to wait.

Ah, that makes sense, the version on my machine runs off the local apt/dpkg database, so I was wondering why why you'd consider it worse than sl. (It does spew several lines of noise, but that's less of a problem when you have a sixty-somthing line terminal and barely a problem at all when you've got scrollback.)

What would you estimate is the ratio of typos versus "thought it was installed"? Personally mine is probably 1,000 to 1. If I type a command and bash says command not found I know the program is not installed. If I need that program I am a big boy and can use apt without anyone holding my hand.

I find `command-not-found``a lot more useful than `sl` (sl is some ascii animation that'll be shown for a short/long while if you write it).

Usually when I get to `command-not-found` it's something I haven't installed on a new system somewhere and I've found it helpful at times. It rarely annoys me.

From what I understand, "sl" was engineered to train new users into typing correctly, by putting them back on track to their final destination, instead of railroading them into promptly installing and running new commands. Only a fascist would force them to run on time.

You can do that too, although I have had an ... interesting experience jumping around desktop environments on Ubuntu. I'm not sure what exactly is happening, but some kind of config information is steadily bleeding between them every time I do a system update.

Sometimes if I pull up KDE (which I don't usually, since I found xfce), the panel will have as many as three different sound menus, only one of which is actually functional.

That's "mechanism, not policy" for you!

From looking at the Gnome Developer page "Desktop files: putting your application in the desktop menus" [1], it seems that it wasn't designed to work correctly with more than one desktop installed. This isn't surprising as it's an uncommon use case. [Edit: This is incorrect, see below]

For example, on a system running 13.04 with MATE installed, the Graphics menu has entries for Image Viewer (Eye of Gnome, the default) and Eye of MATE Image Viewer (MATE's port of eog).

Looking in /usr/share/applications there are the two corresponding .desktop files:

/usr/share/applications/eog.desktop /usr/share/applications/eom.desktop

[Edit]: Neither file uses the OnlyShowIn field which is an extension to the desktop entry format designed specifically to handle this case. [2]

[1] https://developer.gnome.org/integration-guide/stable/desktop...

[2] http://standards.freedesktop.org/menu-spec/latest/ar01s03.ht...

The move to Gnome 3 and Gnome Shell was not Canonical's idea, it was the Gnome Project who decided their roadmap. Canonical in fact developed Unity because of concerns over Gnome Shell. Personally, I rather like Gnome Shell, especially with a judicious choice of extensions, but I have to admit my Core Duo 2 laptops do not always like a fully composited desktop.

Gnome 2 is obsolete, but lives on in RHEL 6 releases, e.g. CentOS/Scientific Linux/Springdale Linux with support until 2017 (apps and hardware) and 2020 (security only). External repositories provide newer kernels[1].

The MATE desktop is a fork of Gnome 2 built on libraries that are being moved forward. External repos for Debian Wheezy and mainstream packages for Debian Jessie. Mint also provide a MATE desktop[2], with the familiar Ubuntu package base.

[1] http://sohcahtoa.org.uk/pages/centos.html

[2] http://linuxmint.com/rel_petra_mate_whatsnew.php

A clarification: The classic look of Gnome 2 sort of lives in RHEL 7 since it will run Gnome 3 with Classic Mode[1] meaning that all that effort for a new UI/UX is wasted as the looks and the "user's workflows" will be the same as it was under RHEL 6 (IOW Gnome 2 like).

[1] http://www.h-online.com/open/news/item/Red-Hat-confirms-GNOM...

I take the point, and it is worth emphasising that the 'classic' mode is the default for RHEL7 installations when you choose Gnome as the desktop environment. However the 'mode' is skin deep - no customisation, can't remove panels, can't add panel apps &c.

I wonder how many EL 7 desktop users will opt for the KDE desktop now that both are presented as equal choices on installation?

I don't think they ever provided you with GNOME2, specifically, they provided you with GNOMEcurrent version, and they still do. You make it sound like they changed their minds, which I don't think is the case.

Are they preventing you from installing GNOME2 on your desktop?

apt-get install gnome-session-fallback

That's now obsolete. With 13.10 gnome-session-fallback is a transitional package. The new package is gnome-session-flashback, so the command is now:

apt-get install gnome-session-flashback

Why are you even using Ubuntu? If you actually know what GNOME2 is and care about it, you are almost certainly not the target audience for Ubuntu. If you use something else, you will probably prefer it over Ubuntu.

Lets examine the popular non-Ubuntu choices that I'm aware of:

SUSE/RHEL/CentOS/Oracle Linux/SciLinux? All RPM based, and while there's nothing wrong with that, and YUM works fine, I already know apt-get and dpkg, why should I switch?

Gentoo? Again, no apt-get. (And no, I don't care if there's a compatibility layer.)

Debian? Wake me up when I can get the latest nvidia driver anywhere in its ecosystem. Meanwhile there's a PPA for Nvidia-$latest that works with my optimus laptop.

Additionally, call it do-release-upgrade, call it apt-get dist-upgrade, call it aptitude full-upgrade. It needs to not cause my system to die in fun and exciting ways. Which it did, going from Squeeze to Wheezy on my server.

Mint? I have yet to be convinced this isn't Ubuntu reimplemented poorly, with less resources, and less attention. How many 3rd party PPAs are there vs Ubuntu? Since I'm digging into using PPAs for bleeding edge releases, why shouldn't I just use Ubuntu?

Linux on the desktop is Ubuntu for me. I threw Openbox on there and I'm good to go with minimal hassle.

If you're passing off Gentoo based on the presence of 'apt-get', you might be failing to understand the point of Gentoo =)

Gentoo uses something called 'portage' -- It's for building everything from source. This gets us access to really bleeding stuff very quickly, while leaving the lid open on an absurd amount of code out there.

This helps me write libraries for the entire Windows and Linux ecosystem. If I can prove that I can support my new binary's ABI there, I've proven that it can survive anywhere.

But if you're looking for apt, you should probably consider any of the binary-based distributions. Gentoo is not for the faint of heart, but if you want to learn Linux and development, it will benefit you like non other.

Debian has apt -- and while its release cycle is a little longer, I've found it to be more stable for mission-critical "apt"/binary-based hosting. In my experience I've found they screw up the packages a little less -- look to how many patches Ubuntu maintains on PHP for a comparison on how dangerous Ubuntu will sometimes play!

Re: Gentoo; I should have complained about 'emerge world' taking days instead. ;)

> it will benefit you like non other.

Tiny point of disagreement; Linux From Scratch is the one true way to learn everything about Linux.

Yeah - I actually run Debian on my servers, but for the desktop, I've yet to find a match for Ubuntu.

Arch? I had it installed on a secondary computer, and after a week I was convinced to switch my main computer over. I haven't looked back since.

Ubuntu was the easiest way for me to get Linux on a new laptop with the secure boot nonsense. And it doesn't involve a lot of tweaking, most stuff works out of the box.

I am typing this on Ubuntu right now (it seemed easier than other options to get it on a Macbook air). I have Manjaro on my other laptop (Majanjaro gets used, Ubuntu is there for the installer), and Mint Debian edition at work (I did think this was a great distro, until a recent update has made it crash a lot).

Well said.

> I will ask members of the Ubuntu community to help to implement this decision efficiently, bringing systemd into both Debian and Ubuntu safely and expeditiously.

Well, this was a pleasant surprise.

I honestly thought that Ubuntu would continue to use Upstart regardless of what Debian decided. It'd have been a monumental amount of work to take on, but on the other hand, there's already been a lot of NIH-syndrome in Ubuntu-land for a while (Upstart, Mir, Unity, etc.)

Ubuntu's been slowly moving in the direction of becoming a "silo", spearate from other Linux distributions for a while now (a direction I dislike). If Ubuntu had decided to double-down on Upstart against the tide, that would have been the final nail in the coffin for them (in my books - they'd still find success for at least a while no matter what they did).

EDIT: Upstart may actually not be a great example of NIH, but Mir certainly is.

One difference is that the init system is not especially user facing.

(It can be, but that means it isn't working very well)

It is for sysadmins. And those are probably Canonical's biggest source of income.

Not really. We're mainly concerned that the interface is consistent.

service ssh restart works fine but that was fine in sysvinit as well..

for some value of fine.

  service my_service restart

required writing 100+ shell scripts, each of which was error prone and difficult.

But also 100% not my problem :)

If one of those scripts forgets to kill a process's children (because the process itself was supposed to handle that), and a server ends up with a bazillion orphaned processes, then it could well become your problem!

This happened to me once. Never again, damn it.

What stopped you from spawning the daemon in a cgroup?

I didn't know about them at the time, and I wasn't the one who wrote the script that spawned it.

It matters to sysadmins and programmers.

I'm assuming the reality that systemd's API will eventually wheedle its way into most other Debian components made the decision for him. Swapping out Gnome for Unity is a lot easier for someone downstream of Debian than continuing to maintain Upstart in the face of Systemd will be.

> but on the other hand, there's already been a lot of NIH-syndrome in Ubuntu-land for a while (Upstart...

Upstart is certainly not an example of NIH-syndrome. At the time that upstart was written, nothing else existed that solved the problems that it did. systemd did not exist at the time. Ubuntu led the way here, and it is unfair to claim otherwise.

launchd existed — albeit was decided against in large part down to licensing concerns (it was, at the time, only available under the Apple Public Source License 2.0 (APSL)), though a couple of months later it was re-licensed under the Apache License 2.0 to alleviate such concerns. OpenSolaris's SMF was considered too. AIUI both were purely decided against on licensing grounds, although they met the technical requirements.

Had launchd already been available under the Apache License 2.0 at the time, who knows, it might now be well-established on Linux. (AIUI, GPL compatibility is not an issue here — the problem is more the APSL isn't compatible with, e.g., the Debian Free Software Guidelines (DFSG) and hence would never be adopted by the Linux community at large.)

unity also predates gnome-shell, I might add.

It does not predate gnome-shell. They were first to call it stable, while GNOME shell was way more stable when we called it stable. You should check the first commits, not when something is released as 1.0. The latter is up to the maintainer and you can do that at any time (and in some cases: never).

Well, the first UNR launcher commit is from 2008-04-16, the first gnome-shell commit is from 2008-10-31. It's pretty close.

I agree with what you said about Ubuntu NIH syndrome.

If their alternatives were better, or at least equivalent, it would not be so bad. But these NIH projects typically only have Canonical (and some Ubuntu) people working on them.

Go to Launchpad bugs and look through open and closed tickets. The NIH projects often cause Ubuntu user problems, and they just lay there and no one fixes them. If Debian, Red Hat, fd.o etc. had other distros depending on things, there would be a lot of people around to fix problems. When Ubuntu does it's own version, they are usually the people who have to fix problems.

> From my perspective the fact that good people were clearly split suggests that either option would work perfectly well.

It's a nice rationality gem right here. If the costs and benefits of both options you're choosing from balance each other out, instead of carefully searching for optimal solution that will yield little marginal utility benefit you may as well toss a coin and be done with it.

Definitely a management perspective.

Sometimes that's the best perspective to have. And in this case, I believe it is. If left up to us engineers/hackers/nerds, we would continue to stubbornly work on or with one system or the other, prolonging the debate indefinitely. And hey, if enough people really feel strongly that Upstart is superior, it's FOSS! Fork and build a community! I know it's not easy but if enough people share your passion it can happen, and that's Yet Another Reason(tm) Open Source rocks.

No, it's an intelligent perspective. I would hope that any engineer would follow similar reasoning and arrive at similar conclusions given the facts.

http://lesswrong.com/lw/th/harder_choices_matter_less/

There's also Piet Hein's classic decision-making strategy when both strategies seem equally good: flip a coin and do what it says. And if you suddenly feel in your gut that the coin picked the wrong option, then you've discovered they weren't equally good.

Actually it's where I first read that insight ;). Though I'm never able to find the link when I need it.

The problem is, of course, knowing both options are roughly equivalent. That may take a lot of work to find out.

It's a tad annoying for us kinda-sys-admins that run full stacks in AWS (or other cloud services) and have to maintain not only domain knowledge for the code we're writing to run on the stack, but the server architecture and services of the OS we choose. My time is limited, and Upstart scripts have been written and run wonderfully.

Ubuntu smooths out many things, which is why I (and I guess many others) choose it for my main server OS. I'm glad to hear 14.04 LTS will still have Upstart support, but this means I have to put "move to systemd" as a medium-range ticket, which is sadly more time wasted.

Go progress! But comon, progress! Stop making more work for me. :)

On the other hand, you'd be feeling quite differently if your AWS stacks currently consisted of RedHat or CentOS (as they oftentimes do for larger/enterprise environments).

I agree, you drew the short stick on this particular transition, but overall, this is a win for sysadmins in the long run, as it seems like all the major distros[0] are moving towards systemd.

Systemd isn't that much more to get used to if you already know sysvinit, and this way your scripts will be more portable to other distros, not less.

[0] IMHO, Gentoo isn't a distro so much as a meta-distro - "create your own distribution".

Not sure about the first part of your comment as neither RHEL nor CentOS use systemd right now (in fact RHEL 6/CentOS 6 uses upstart).

RHEL 7 will be systemd based, but it isn't ready (yet), and in my experience people tend to stay in the major release they're in instead of upgrading. So it will be a long time before systemd is really widespread between RHEL/CentOS users.

RHEL6 was released in 2010. That's why they are "still" using Upstart, because systemd didn't exist when RHEL6 was being qualified. Fedora (which is upstream/testing for RHEL/CentOS) switched to systemd years ago. 3 releases ago at least. RHEL is the equivalent of what Fedora was a few years ago, plus testing and bugfixing by people that get paid from the billions in support contract money that Red Hat pulls in every year. RHEL7 betas and rcs have been available to server admins under contract for a while, they are likely figuring out upgrade issues currently, and when it comes out they probably will switch over a lot of their machines. All those people have support contracts with Red Hat, the packages are qualified and if something goes wrong they get to call up and complain. If you are coming from the Ubuntu world RHEL is Long Term Stable, and upgrading to a new LTS might be work for your admins but the OS should be solid already.

You have all of what, 5 years to transition away from upstart before 14.04 is no longer supported? Medium-range indeed.

I'll nestle it here, just after my still-open 'Y2K preparation' ticket.

You host most of your stuff on a public cloud which means you don't do a lot of what most sysadmins do. Do yourself a favor and embrace it. Learning new tech and (more importantly) being able to know from experience which is more fit for a specific task is part of the career field you've chosen.

I've professionally managed thousands of RHEL/CentOS/Fedora/Gentoo/Ubuntu/Debian servers in production (and that is just linux!) and not had any real issues learning the variations. Sure you learn different package managers, /etc/sysconfig vs /etc/default, different init scripts, but it is just semantics. That is why sysadmins get paid, to be experts that know the differences.

It is a good thing to know both, embrace the new tech.

EDIT: Oh and I gave you an upvote, this isn't mean to be a snark.

I don't see any difference in volume and complexity between sysadmin work on physical iron and public cloud systems.

Everything you are describing applies to VPS systems.

There are differences, but those have more to do with optimal use of resources, and in that aspect managing cloud systems is actually more complicated. At least the behavior of your own physical systems is consistent and predictable.

Besides touching physical machines, what is it you think a sysadmin for cloud systems doesn't do?

The overwhelming majority of layer 1 monitoring and / or troubleshooting. Using a cloud provider, by design, outsources all of the layer 1 monitoring and troubleshooting to the cloud service provider. If you're not familar with the OSI Layer Model[1], layer 1 == the physical layer ie: hardware.

How often do you need to monitor your EC2 instance for failed physical disks or bad sticks of memory... never :)

Additionally, it depends on how much you outsource to your cloud hosting provider. Many people trust ELB to do the right thing. It is a pretty reasonable product, but having used keepalived + the default builtin LVS[2] that is in every distro kernel out there for more than 10 years, why trust Amazon to do it better? Why use Amazon Dynamodb when there is Riak[3] or Amazon Elastic Cache when groupcache[4] is so awesome?

Using cloud services == outsourcing some/all of the traditional sysadmin work. Many people who consider themselves devops people are smart developers with very little real sysadmin experience. Please don't fall into the trap of considering them exactly the same. It is an interesting hybrid role, but is distinctly different.

[1] http://en.wikipedia.org/wiki/OSI_model

[2] http://www.linuxvirtualserver.org

[3] http://basho.com/riak/

[4] https://github.com/golang/groupcache

> Learning new tech and (more importantly) being able to know from experience which is more fit for a specific task is part of the career field you've chosen.

Yes, but that doesn't mean that any particular instance of having to learn new tech is something ve should celebrate. The opportunity cost of learning this might be that ve doesn't learn something else useful.

Yes with Upstart! Now if only AWS can come up with scheduled starts and stops, the most asked feature (but why stop server why you can keep making money off idle resources?). Then my life is complete.

Check out Ylastic for this - http://ylastic.com/Features-Sheduling.html.

Thanks, shirke. My current setup is using a AWS EC2 Micro-instance (that I set up on free tier) as a scheduler to start and stop my other instances. Ylastic looks cool but I make it a point to never pay for SaaS, Others who are interested can google the term 'Boto AWS scheduled start stop.'

Kudos to him for this. Could not have been an easy decision but does sound like he listens and wants what is best for Ubuntu in the long run.

It's interesting he doesn't mention that at least one significant factor against Upstart (from the perspective of Debian) was the Canonical Contributor License Agreement.

http://www.youtube.com/watch?v=CUOAUQJ-y00&t=42m15s

That was interesting? You expected him to shine a light on their "icky problem"? Every community has their "icky problem" that they do not like to talk about in mixed company, if at all. The CCLA is Canonical's icky problem.

Wow, i wouldn't have thought that. Since i moved away from Ubuntu to Debian mostly because all this upstart, mir, advertising solo attempts, it's nice to hear that.

Now i'm waiting for the same blog post about supporting Wayland and i may even switch back sometimes.

Who is using Wayland right now?

Who is using Mir right now?

Not even Ubuntu. They use Mir only as a wrapper around XMir which then displays Unity – they've won nothing with this.

But with Wayland coming to kubuntu this summer and Mir coming to Ubuntu at the same time, we'll see how Wayland and Mir compare in production.

Mir has been shipping on Ubuntu Touch since 13.10.

Well that begs for another question, I would say...

Indeed, I think we're about to see some very interesting times in the next year or so, with regard to GUI/Windowing systems in the Linux world.

Wayland has been a long time coming, and I do hope that's because those working on it are really shooting to get things right. Mir is interesting, and it looks like it will be able to do some things that Wayland/Weston can't, or would require some difficult hacks or even spec changes to make work. I don't know if Mir would be better than Wayland for me, but if you ignore all the FUD from both sides, I think the bottom line is that Mir is better for Canonical's plans unless or until Wayland/Weston can be changed to accommodate their needs.

I personally have no positive or negative opinion on Mir vs. Wayland, but I hope that at least one truly delivers on their potential and shows a clear improvement over our old trusty ;) X11/Xorg.

AFAIK, Packages are available in Arch and the latest releases of Fedora may have Wayland packages available, but it's not (yet) the default window system for either distro. Fedora has been planning to ship Wayland in the main distro for some time, but it's not an easy task, and it keeps getting pushed out to "the next release". They want to do things right, and seem to be the biggest supporter of Wayland, so it may still be some time before a major distro starts using it as their primary GUI layer (aka windowing system, though I'm not so sure that's an accurate term for these things anymore)

In addition, GNOME is busily porting everything to work on Wayland. I think when that is finally done, we will begin to actually get Wayland-by-default on a few dists, and from there, well... that depends on how well it works!

Links to additional info:

  https://fedoraproject.org/wiki/Changes/Wayland
  http://blogs.gnome.org/uraeus/2013/09/09/fedora-wayland-update/
  http://fedoramagazine.org/?p=538
  https://wiki.gnome.org/Initiatives/Wayland
  https://wiki.gnome.org/ThreePointEleven/Features/WaylandSupport

I personally have nothing against either Wayland or Mir, I just want my system to work and if I can get better security, performance, features, etc out of it... score!

Wayland is used in the Sailfish OS shipped on Jolla phones. It also seem to have traction on Raspberry Pi, mostly because X really sucks on their low-powered hardware.

Gnome is likely to be the first DE to run on wayland. Gnome 3.12 is being delayed to release lockstep with wayland - especially to support some of the gnome-settings-daemon stuff.

Wayland packages are in Ubuntu already and will likely be used by desktop environments other than Unity.

Losing graciously is always deserving of respect. Thank you Mark.

s/upstart/Mir/g

s/systemd/Wayland/g

s/2014/2015/g

Your regexps are wrong, these replacements don't make any sense at all.

Unless there was an edit, they do actually make sense, if you think of this post as being rewritten in a year to discuss a move away from Mir (current Ubuntu graphics thing) to Wayland (current non-Ubuntu graphics thing that most everybody else is probably moving to).

Right now they say replace all occurrences of "upstart" with "Mir" and all occurrences of "systemd" with "Wayland". I can't see how that would work.

"Ubuntu is switching from upstart to systemd in 2014" -> "Ubuntu is switching from Mir to Wayland in 2015"

Why would they drop Mir? Dropping Upstart makes sense because systemd has momentum and they both do basically the same job. With Mir vs. Wayland they already evaluated Wayland and decided it wasn't going to be suitable for their needs, unless Canonical change their target markets this isn't something that'd make sense to change (unless a later version of Wayland supports what they want).

It is the common opinion that their needs (re Wayland/Mir) are not technical, but related to defining and shipping the results product resulting of those efforts.

That's how it appears, but Canonical wouldn't duplicate work unless they thought it was in their best interests, the question then becomes what needs are they trying to meet? Perhaps the needs are technical, perhaps not.

I understand part of the issue was Canonical changing their support for Wayland, but what exactly does Mir do to make developers lives harder? Most developers don't write for X directly, nor will most developers write for Wayland or Mir directly, so as long as the library support is in place then why should we care? I don't hear Android developers complaining about Android using its own window manager. If Canonical want to pay extra for their own window manager, and are prepared to work on the libraries that make developers lives easier, what are we really losing?

"That's how it appears, but Canonical wouldn't duplicate work unless they thought it was in their best interests"

Yes, their best interests. Not to the Linux ecosystem.

Maybe if they do what's in their best interest, the best things happens for the Linux/nix ecosystem. I'm not exactly sure what your arguing about; All corporate entities seem to (perhaps quite reasonable considering they're corporations) cater to their* own interests first.

From what I've gathered, we have, as a community gained a lot from that by looking at Red Hat Inc, Canonical Ltd, SuSE GmbH/Novell/Attachmate Group and others contributions.

Give me an example where Red Hat or SuSE have started a brand new project that directly competes with an existing effort with broad support from the open source community?

Best example is probably .deb (1995) vs. .rpm (1997).

I am wondering if they are ever going to do this with bzr. I did not realize how much bzr development had slowed until I read the emacs bzr/git debate.

Now that's a surprise. I've never known Canonical to give up like that.

There's still time for Shuttleworth to get pissy yet.

In his own title, he called himself gracious, which struck me as what people do when the graciousness comes very hard to them and they have to make a strong conscious effort.

So possibly he felt pissy about, but rose above that.

I think he'd get extra graciousness points if he hadn't called himself gracious, though.

My thought as well. It's reminiscent of people listing humility as one of their qualities.

"(today our focus is on the cloud and on mobile, and we are quite clearly leading GNU/Linux on both fronts) "

How is targeting two widely-different platforms focus?

And how exactly is Ubuntu leading Linux in the cloud?

Pretty much every source I can find shows Ubuntu with a clear lead, if not an actual majority in marketshare on EC2 instances.

Canonical are making a bet on the graphical subsystem re Mir/QT/QML/Oneclick to crack touch/mobile. Similar to Apple going all-in on Objective C/Cocoa(Touch).

But for the low level non-graphical parts of Ubuntu it makes a lot of sense to stick to standard Linux/Debian. At the terminal/sysadmin/command line level a standard setup is desirable.

The headline made me think he was going to talk about losing the desktop and mobile wars!

I'm really sad about this. I love Upstart and think it's an elegant init process implementation. Anyone more informed than myself care to summarize the key points in making the decision?

The summary is essentially just that systemd is just a bit more elegant. Basically the (non gamed) votes had those two at the top, but in the end upstart had a few more pain points, and systemd had a few more features.

- SystemD parallelization is more automatic. - SystemD has built in logging for all managed processes. - SystemD has Socket Activation (start process on demand) - SystemD has more momentum in other major distros (Fedora, Arch) - etc..

So when will we see them drop Mir?

When all other distros start using Wayland, so never.

Mir hasnt lost the race, it just entered late. You could critisize them for entering late, but in the end it doesn't matter. As long as people are working on viable X alternatives.

I am an Ubuntu member and I was deeply wondering what I was going to do next... Tollef had me realizing systemd is the future for quite a while now :-)

Mark reminds us once again what the mission of Ubuntu is, why it is the best alternative for end-user computing in the FLOSS world and, for me personally, that I have a home already and do not need to go anywhere.

And... onward to build on top of the great foundation we already have.

Nice move indeed! Let's just hope that there will always be viable alternatives, just in case the systemd project ends up like pulseaudio.

How did pulseaudio ended up? It seems to me that it's quite a healthy project right now.

It may have had some issues at the beginning but I can't find anything to complain about it right now.

The flipside is that diversity in software is usually mentioned as a Good Thing. Losing Upstart means losing diversity, no?

Upstart although being first was never going to win against the marketing forces behind systemd. In the old days, tech was evaluated on technical merit, today the slickest marketing hucksters win. Yet another piece of lennart-ware to bring down the linux experience. Thank you BSDs for providing a sane alternative.

Canonical, and more personally, Mark have been targets of my frustration with the Linux desktop, I moved like many others to Mint. This is a very welcome reminder that even if phrased negatively as 'lessor evils', Canonical contributes and attracts people from various backgrounds to something I sincerely want to overtake the current selection of OS options (more philosophically than anything). Canonical may not listen very well, and ultimately that may hurt any chances of desktop success for linux, however this reminds me they're still in a very small circle of 'lessor' evils.

The "new stewards of PID 1" == welcome to your imminent doom.

Graciously?

What's not gracious about that post? Losing graciously doesn't mean giving up without a fight, it means once you've lost, you take it well.

I actually found it somewhat less than gracious. I read it to say that the wrong decision has been made, but accepted - an attempt to gain political points for the next round.

If you have to say you're being gracious, you're probably not.

It's clear that he thought Upstart was a better choice. His opinion hasn't changed, but he accepts that the majority of people who voted on this thought differently, and accept their decision.

If you're convinced something is good/better, and you've argued that way for months, do you change your mind about all that when a committee or manager decides to go the other direction anyway?

> I read it to say that the wrong decision has been made, but accepted

Sorry I don't read it your way, he seems to recognize that both systems would make an excellent choice:

"From my perspective the fact that good people were clearly split suggests that either option would work perfectly well."

Being gracious in this sort of instance doesn't necessitate admitting you were wrong.

I graciously admit that is correct, but not my objection.

If you make a big deal about being gracious, that's the opposite of gracious.

To certain values of 'gracious.' I think part of graciousness as a winner is not to ask the loser to humiliate themselves for you.

If he didn't think that Upstart was a better choice yet still argued for it, that would mean that he had been wasting all of our time.

You're confusing grace with humility. There's no problem with people referring to their graciousness, as long as they remain courteous while doing so.

Anyway, 'big deal about being gracious'? Your second sentence contains more instances of the word than Shuttleworth's entire missive. The title mentions grace; the text says "we ran hard, we lost, we'll go with our upstream distro on this one, thanks everyone". The article is essentially a wake for Upstart, a near-10-year project initiated and championed by Canonical, and Shuttleworth is thanking everyone for their past work and work yet to come.

Ironically, you're being very ungracious here.

Er...

From my perspective the fact that good people were clearly split suggests that either option would work perfectly well. I trust the new stewards of pid 1 will take that responsibility as seriously as the Upstart team has done, and be as pleasant to work with. And… onward.

Looks pretty gracious to me.

I'd guess that, given Canonical/Ubuntu has struck out on its own in various other respects (and the level of controversy that generates on forums like this one), there is a need to make clear to the broader community that they aren't doing that on this particular issue.

(Speaking as a naive desktop ubuntu user who wouldn't know systemd from upstart until they blow up in my face. Which, AFAICT, neither ever has.)

> If you have to say you're being gracious, you're probably not.

This is the little brother of "You know I am not racist, but..."

> Graciously?

Question mark much without explaining anything?

You were too lazy to post more information or what exactly you thought was not gracious about it.

It seems very gracious to me. But if you're not convinced, compare it to Ian Jackson, who is seriously pushing for GNOME to be removed from Debian if it has dependencies on systemd:

https://lists.debian.org/debian-ctte/2014/02/msg00485.html

Actually, now that I think about it, this announcement has rather serious implications for the current debate in Debian. Several members of the Technical Committee are proceeding under the assumption that the Upstart developers will continue to provide logind without systemd. If Ubuntu is moving to systemd, that work is a dead-end, and Debian is going to have to take over maintenance of that work if they want to use it to support multiple init systems.

The decision was already made, unless they can approve _and_ win a GR for overriding it (and I don't think they would win).

The decision to make systemd the default in jessie has been made. The TC is still debating whether or not packages will be able to declare systemd as a dependency.

They will continue to provide it. Part of what he wrote: > We’ll certainly complete work to make the new logind work without systemd as pid 1.

I don't think that is a fair reading of that email. The remark at the bottom about preserving choice is in line with what Jackson has said before.

Here's the last paragraph:

"The doomsday scenario of choosing between (a) and (b) becomes less likely if we make it clear how bad it would be. We need to provide appropriate backpressure to encourage upstream decisions that support the continued freedom of our users."

In other words, Jackson is willing to drop GNOME from Debian to pressure GNOME developers into not making dependencies on systemd.

No, and beyond being inaccurate I'm beginning to think you are being intentionally misleading.

He wants the doomsday scenario of choosing between (a) and (b) NOT to occur. By making it clear how bad it would be.

But his "doomsday scenario," where Debian has to drop support for all other inits or drop all support for GNOME, is a false dilemma. Debian can still support other init systems, for people who want them, while still letting GNOME depend on systemd if that's what's best for GNOME. Users who don't want systemd can run without a DE, or they can run a DE that doesn't require systemd. (GNOME isn't even the default DE for Jessie, that's Xfce.)

So yes, he wants to make it clear how bad it would be, but he's also the one who's pushing for making it bad. It's not like he's just warning us that someone else will make Debian choose between those two things, he's trying to make Debian choose between them.

I think that's the right choice. GNOME is leaving non-linux out in the cold. There are plenty of linux distributions that ship GNOME, but there's only one serious GNU/kFreeBSD distribution.

but is there even one serious GNU/kFreeBSD user?

There never will be with an attitude like that.

I use Debian on my laptop, and FreeBSD on my server, but I wouldn't touch GNU/kFreeBSD with a ten foot pole. FreeBSD kernel with GNU userland and SysV filesystem layout is the worst of both worlds.

I'd rather see the inverse: a Linux kernel with BSD userland, filesystem, and ports. But that doesn't fit with Debian's strengths as a package repository. Installing packages to different paths on different systems is not really feasible. Using ports means Debian has nothing to contribute, since it completely replaces the Debian package system.

On the other hand, swapping out just the kernel is a natural fit for their system. All they have is a hammer, so they found a nail. Unfortunately it doesn't result in a very appealing OS, since the kernel is not really the selling point of BSD. It works, but it's nothing special. The magic is in the BSD userland.

I like the solidity of BSD and I love ZFS, but the ports system always felt surprisingly flimsy (coming from Gentoo I was expecting it to be just like portage, but it's a lot more ad-hoc and breakable) and I never got on with the BSD userland - I mostly work with alias ls=gls, cp=gcp and so on. If I set up my server again it would probably run GNU/kFreeBSD.

Ugh. While I agree with some of his concerns about the ever-expanding scope of systemd, all of what he has managed to do during this debate is paint himself as an immature, toxic person, between his call to remove the head of the Debian technical committee and his "ultimatum".

Reply on news.ycombinator.com