Distribution quotes of the week

[Posted November 28, 2012 by ris]

In a way Canonical is just doing something that Android has pioneered. Take the kernel, share some userspace components, but turn it all into your own private platform. That's certainly a valid strategy. And they might even pull it off (in contrast to Google, they create the platform gradually instead of abruptly) But I'll leave it to you to figure out what this all means for the Linux platform and ecosystem in general...

-- Lennart Poettering

Yes it's a valid strategy for Ubuntu, but it's not really a friendly one is it? Taking the rest of the Linux ecosystem into account that is. I remember when Ubuntu had the tagline "Linux for Human Beings". Now I'm starting to doubt they want to be associated with Linux.

-- Andrew Powell (in the comments)

One of the nice effects of open source is that there are absolute limits to crazyness. People are free to do whatever crazy stuff they want. Someone else will step forward and fork it back to sanity if necessary. This system works very well.

-- Bjørn Mork

to post comments

Distribution quotes of the week

Posted Nov 29, 2012 4:05 UTC (Thu) by rsidd (subscriber, #2582) [Link] (23 responses)

So "the platform" is whatever Lennart/Fedora want it to be? Upstart was there first, as one of the commenters says. Neither upstart nor systemd is part of "traditional" Linux. And there are those who say the "platform" should be Unix-like systems, not Linux and certainly not Fedora specifically. Having abandoned the BSDs and others (Lennart is quite open about not caring about them), the next step is to abandon Ubuntu, Gentoo and others. Fine, but don't pretend that you are supporting some sort of "standard platform" and they are not.

Distribution quotes of the week

Posted Nov 29, 2012 7:21 UTC (Thu) by rahulsundaram (subscriber, #21946) [Link] (22 responses)

You are overreaching in your criticism. systemd works on Ubuntu and Gentoo just fine. If the distributions don't choose to make it the default, that is certainly a valid choice but Lennart didn't abandon anything in that case. Moreover, if you read the full blog post, Lennart doesn't claim systemd is a standard platform. Quite the contrary, he says he has failed to achieve the standardization he set out as one of the goals of systemd, in part because systemd developers haven't pushed for use of it to manage user sessions yet although it is possible to do so now.

Distribution quotes of the week

Posted Nov 29, 2012 9:35 UTC (Thu) by rsidd (subscriber, #2582) [Link] (21 responses)

Lennart wrote:

Whichever way you turn this, with this move Ubuntu will have their own platform on most levels now. They share the kernel, but not the system manager, session manager or UI (Unity...) with Fedora, Suse, Mandriva, ArchLinux (and other systemd using systems) anymore, i.e. the cut goes through the entire userspace. The separation of the userspace bits of the OS is not complete, we do share components (such as D-Bus, ...), but this is precisely where the focus appears to be now: sharing components of the platform, rather than the platform itself.

The only way I can interpret this is, he thinks systemd is "the platform itself" and not an (optional) "component of the platform", and Ubuntu, by not adopting it, is opting for their own platform. And I really don't see why the session manager or UI are part of a "platform definition". Have we totally stopped catering to those who launch "startx" from a console to enter an FVWM session?

Distribution quotes of the week

Posted Nov 29, 2012 10:40 UTC (Thu) by cesarb (subscriber, #6266) [Link] (8 responses)

The way I read that is that the "platform" Lennart meant is the whole stack, from the desktop environment (Gnome, KDE, etc) down to the kernel. Notice he also mentioned Unity.

Distribution quotes of the week

Posted Nov 29, 2012 12:27 UTC (Thu) by dgm (subscriber, #49227) [Link] (7 responses)

It's not a problem with stacks, but with numbers. Lennart implies that being used by more distros means that systemd is "more standard" (my words, not his) than Upstart. Another completely fair argument would be that the number of distros is irrelevant, and instead it's the number or users what counts, and thus Ubuntu probably wins.

Because there's no mention of any option of sharing interface between both Upstart and systemd, I conclude that this feature will cause a fracture between the two contenders. And apparently both are to blame for that. It's like GNOME vs KDE all over again, only on init space.

The Linux world is full of such fractures: RPM vs deb vs anything else is just another example.

Distribution quotes of the week

Posted Nov 29, 2012 14:37 UTC (Thu) by rahulsundaram (subscriber, #21946) [Link] (6 responses)

If you are counting number of users in the desktop maybe but if you are counting in other segments, RHEL and rebuilds like CentOS are used far far more and if you round up all the distros using systemd as default now and in their next releases, including RHEL 7, systemd wins.

Strategically speaking, what the major distros are using is very important since development of other components are going to be affected by that and unless you have the resources to manage everything on your own, you should seriously look into aligning with them unless the cost of moving over is too high and in that case, interface compatibility would be a good thing to shoot for. This goes beyond just systemd of course as hinted in the blog post.

Distribution quotes of the week

Posted Nov 29, 2012 17:02 UTC (Thu) by dlang (guest, #313) [Link] (1 responses)

This is assuming that RHEL7 actually ships with systemd, and that everyone who uses older versions upgrades.

Given the huge volume of users who are still running RHEL5 or older, even almost two years after RHEL6 has come out, counting all RHEL6 and earlier users as 'systemd users' is misleading at best.

As Microsoft has shown us, just because a company has a huge userbase and has released a new version of their OS doesn't mean that that new OS will become the default everywhere, let alone the default anytime soon.

Distribution quotes of the week

Posted Nov 29, 2012 20:23 UTC (Thu) by rahulsundaram (subscriber, #21946) [Link]

RHEL 7 coming out with systemd by default isn't a assumption at this point. I wouldn't count all RHEL users but rebuilds do get deployed on newer systems faster than you think but I wasn't asking to count all or any EL users as systemd users. What I am pointing out is the strategic importance of systemd beyond just a mere user count.

Distribution quotes of the week

Posted Dec 2, 2012 21:40 UTC (Sun) by tpo (subscriber, #25713) [Link] (3 responses)

> but if you are counting in other segments, RHEL and rebuilds like
> CentOS are used far far more

Can you show me the numbers and where they come from? If you have them, then it'd be nice if they were clear about whether they are counted under the assumption that the world ends at US borders.
*t

PS: in my universe most people are using Debian as servers. *Some* very large companies use RH/CentOS, but it's small companies that generate the most revenue, so they possibly comprise the most servers. Smartphones don't use RH. Appliances I don't know, I know some specific ones that do use Debian.

Distribution quotes of the week

Posted Dec 3, 2012 6:05 UTC (Mon) by rahulsundaram (subscriber, #21946) [Link] (2 responses)

One could accurately tell you the number of active subscriptions that RHEL has but that is internally information for Red Hat and similarly for SUSE. Other than that, no accurate numbers exist for any distribution typically. That are ballpark figures for some community distributions. On the desktop level, Ubuntu is clearly popular but they clearly don't have the lead on the enterprise or server level

* Linux mirrors and download stats from them
* What webhosting administration panel supports including cpanel
* What do the most popular and most widely used webhosting companies offer by default
* What do server ISV's support the most

Do your own research and draw your own conclusions.

Distribution quotes of the week

Posted Dec 3, 2012 7:03 UTC (Mon) by rsidd (subscriber, #2582) [Link] (1 responses)

None of this is in the least relevant! The question is whether the biggest distribution (be it RHEL, Fedora, Ubuntu, or whoever) is allowed, de facto, to "define" the platform. If one distro commanded 90% of the users, clearly that would win, but we aren't there yet. Lennart's suggestion that Ubuntu, by not going along with his vision, is not sharing "the platform", is obnoxious, as is defining "the platform" to include gdm and gnome3.

Below, tpo writes: 'If the biggest desktop distro says "we chose these technologies [systemd, gnome3 etc], and we'll be using them during the next 10 years and we guarantee they will be around during that period", then that'd be different [for proprietary software].' That's what Fedora and Red Hat are aiming at. Best case (for them) -- other distros fall in line. More likely -- proprietary software distributors will continue to package their stuff separately for each distro, but the differences will not just be about dependencies and packaging format, but will involve basic parts of the system.

I'm not a Fedora or RHEL user and am not convinced I will ever want to be one. Not only am I repelled by this needless re-architecting of Unix methods that have worked for decades, but Gnome3 and Unity are so competitively awful that I have just switched to a tiling WM (i3) and don't plan to go back. In this mad and futile rush to win the "desktop" (which was last year's battle -- the world has moved on to mobile devices), RH/Fedora and Ubuntu have both managed to annoy users who just want a working Unix-like system whose parts fit together in easily comprehensible ways and whose configuration can be changed with a text editor.

Distribution quotes of the week

Posted Dec 3, 2012 7:18 UTC (Mon) by dlang (guest, #313) [Link]

that may be what RHEL is aiming for, but Fedora is not

At least not unless they are willing to give up the 'leading edge' goal as well, or unless they are so arrogant that they thing that the decisions they will make now are going to be so good that no better decisions will be able to be made in the next 10 years.

Distribution quotes of the week

Posted Nov 29, 2012 18:20 UTC (Thu) by jspaleta (subscriber, #50639) [Link] (10 responses)

You don't think the UI is part of a platform definition?

So for iOS the UI and the associated API it exposes is not part of the "platform" defintion? For osX? For win8 the UI and the associated API it exposes is not part of the "platform"?
For Android.....
For Maemo....
For WebOS....

I think the exact opposite. I think right now, to be competitive in the marketplace as a platform for a range of consumer oriented computing devices that are meant to be extendable with add-on application software.... the "platform" definition is generally expected to include the UI to build a bridge to those application developers.

I believe that "Canonical" is opting to build a distinctive platform..yes.
I believe there are compelling strategic business realities which make that sort of differentiation inevitable really. OEMs want a partner, and that partner needs to pitch a distinctive platform to be compelling to those OEMs.

I think its inaccurate to claim that "Ubuntu: The project" or "Ubuntu: The community" is attempting to build a platform. The platform building effort is very much a Canonical strategic effort. "Ubuntu: The project" and "Ubuntu: The community" to a large degree are trying to maximize what is offered under the project resource constraints. The fact that we are talking about "Ubuntu" building a platform at all is just the side-effect of the choice Canonical has made to intermix community/project and product interests under a common brand. I'm somewhat hopeful that the Unity brand becomes stronger and that they start pushing the Unity brand as the platform brand instead of overloading the Ubuntu brand. Calling it the "Unity Platform" is much cleaner strategically than calling what they are building the "Ubuntu Platform."

That being said, I think the Unity platform, would be made stronger in the marketplace if Ubuntu adopted systemd, even at this late date.

-jef

Distribution quotes of the week

Posted Nov 29, 2012 19:40 UTC (Thu) by el_presidente (guest, #87621) [Link] (1 responses)

> You don't think the UI is part of a platform definition?

It's the blog post that mentions a "linux platform". Since Linux doesn't have a standard UI, it's not clear what Lennart is referring to.

Distribution quotes of the week

Posted Nov 29, 2012 20:39 UTC (Thu) by jspaleta (subscriber, #50639) [Link]

Linux as a platform does not yet exist. "Linux: the platform" is something being worked toward slowly.

"Linux: the kernel" is a technology building block that can be used as part of a "platform." "Android: the platform" is one such construction. "WebOS: the platform" was another. There have been others as well, its clear to me at least that Canonical is moving toward defining a platform where "linux: the kernel" es muy importante.

Can "we" find enough consensus to be able to define "Linux: the platform?"

I'm not sure we can. Platforms are not least common denominator sorts of things. More realistically we are going to see a family of strongly branded platforms, with strong but diverse API policies. Each leveraging "linux: the kernel" and each legitimately being able to claim being a "linux platform" but not necessarily "the linux platform". making linux a very weak platform brand.

We would be better served right now if we admitted that, encouraged the use of distinct platform branding, while simultaneously encouraging and agitating for standardization on plumbing level components which sit above the kernel across those platforms.

The real question is how far up the stack can we get commonality? I think systemd brings enough benefits to the table to be part of the shared stack of platforms which make use of "linux: the kernel". It might be disruptive for some, but significant technology improvements tend to be.

-jef

Distribution quotes of the week

Posted Nov 30, 2012 4:26 UTC (Fri) by rsidd (subscriber, #2582) [Link] (7 responses)

You don't think the UI is part of a platform definition?

Indeed I don't, at least not on Linux. There has never been a standard UI. Even today, on a daily basis I use programs that use xlib, motif/lesstif, tcl/tk, as well as GTK and Qt. And in what sense is Unity a "platform"? It depends on Gtk just as much as GNOME3 does, and all the same software runs on it. And if Ubuntu uses lightdm instead of gdm, who cares? How does it affect the platform?

The linux platform is a system on which you can run the linux versions of Adobe Reader, Skype, Opera, Mathematica, Matlab, Google Earth, and various other proprietary software in addition to the usual open-source software. It exists and it works, and GDM/GNOME3/lightdm/unity/systemd/upstart are not essential parts of it. Defining a "Linux platform" that includes systemd or GNOME3 will not increase the availability of proprietary software, either.

Distribution quotes of the week

Posted Dec 2, 2012 21:57 UTC (Sun) by tpo (subscriber, #25713) [Link] (6 responses)

> Defining a "Linux platform" that includes systemd or GNOME3 will
> not increase the availability of proprietary software, either.

I'm not sure. If "defining" means some random persons decrees it, then no, it probably won't change anything.

If the biggest desktop distro says "we chose these technologies, and we'll be using them during the next 10 years and we guarantee they will be around during that period", then that'd be different.

When you write desktop SW, there's some choices you have to make:

* widget set?
* indicators?
* desktop integration (desktop tray)?
* configuration management?
* display/editing of files based on ... mime?
* setting up services?
* etc.

Now if you know you have a stable platform to program against that will be able to reach the majority of your users for the next 10 years, than that's a big bonus for you (the developper/company).

Distribution quotes of the week

Posted Dec 6, 2012 9:56 UTC (Thu) by yeti-dn (guest, #46560) [Link] (5 responses)

> If the biggest desktop distro says "we chose these technologies, and we'll
> be using them during the next 10 years and we guarantee they will be
> around during that period", then that'd be

an utter lie.

To give you a brief idea what 10 years means, 10 years ago Fedora did not exist yet. Neither did Firefox, cups, udev, Xorg X server, ...

The only thing guaranteed by such promise would be that the desktop distro itself will no longer exist in 10 years (should they choose to stick to it).

Distribution quotes of the week

Posted Dec 6, 2012 10:10 UTC (Thu) by rsidd (subscriber, #2582) [Link] (4 responses)

And yet, Ubuntu Dapper was released in June 2006 and supported for five years as promised. It does not seem impossible to me that a 10 year guarantee of stable technologies would work. In a sense it already does -- binaries written in 2002 (or 1992) are quite usable today provided you have the compatibility libraries. To claim that systemd and Gnome3 will be supported as "platforms" for 10 years means that vendor will supply compatible versions of those things on all releases for ten years, even if nobody else does and even if a totally incompatible Gnome4 is released in 2014. It can be done.

Don't use phrases like "utter lie" lightly.

Distribution quotes of the week

Posted Dec 6, 2012 10:26 UTC (Thu) by dlang (guest, #313) [Link]

go back and look at what was supported in Dapper for those 5 years.

I believe that it didn't include 'desktop' things, only 'server' things

also look at how much (or how little) is supported in the RHEL2/3/4 releases over time.

some parts of the stack are pretty stable, these include the kernel, webserve, and sysV init.

other parts are very unstable, and maintinance of 'frozen' ports is not practical. These include Firefox, Gnome, and probably, given it's current rate of change and development, systemd init and udev.

Distribution quotes of the week

Posted Dec 6, 2012 10:31 UTC (Thu) by tpo (subscriber, #25713) [Link] (2 responses)

I'm not sure if yeti-dn's comment is more than trolling. You rsidd mentioned Ubuntu's 5 year guarantee. I'd add Red Hat's [1]:

"Red Hat Enterprise Linux 5 and 6 are offered with 10 years of Production Phase support, followed by a three year Extended Life Phase."

[1] https://access.redhat.com/support/policy/updates/errata/

Distribution quotes of the week

Posted Dec 9, 2012 10:34 UTC (Sun) by yeti-dn (guest, #46560) [Link] (1 responses)

Five years are about the edge for a desktop system. Both in the sense that it is doable since things are starting to fall apart but the technologies have not diverged too far yet and it may be still acceptable for conservative users.

Please go and read the actual conditions for RHEL. The later stages and extended support are clearly meant to just keep installed systems running. You get no new features, no support for newer hardware, services, file formats, etc. This is fine for a server that has a dedicated task to do. It is also fine for a single-purpose user machine, for instance used to control an instrument where you do not want, ideally, any changes at all (hence, little support). But this discussion started about desktop distros. And for a desktop (as in general purpose user system), not being able to connect most hardware newer than 10 years old, open any file formats that have appeared or changed substantially in last 10 years, use various services due to too ancient clients, ..., is deadly.

Also note you pay rather a lot for RHEL support -- fixing security holes in ten years old programs is terribly ungrateful work. Since community distros cannot make the contributors do it, those that tried to introduce this kind of long term support have all failed.

Distribution quotes of the week

Posted Dec 9, 2012 17:46 UTC (Sun) by raven667 (subscriber, #5198) [Link]

But let's be clear, the extended support kicks in at the end of life, a life that does include feature updates so you'd have to count from the end of feature updates, not from the beginning ship date. Also, CentOS for those who aren't going to pay. In any event, as you say, the refresh rate for desktops is 5yrs or so which would bring new versions of the supported enterprise-style OS, one like RHEL which is highly backwards compatible with previous releases

Distribution quotes of the week

Posted Nov 30, 2012 7:01 UTC (Fri) by apoelstra (subscriber, #75205) [Link]

> Have we totally stopped catering to those who launch "startx" from a console to enter an FVWM session?

I do this, and I use systemd. (In fact, it is was much easier to find where systemd was starting X, and telling it not to do that, than it often was with init. If I was lucky, I could set the default runlevel to 3. Otherwise I'd be reading through /etc/inittab, /etc/init.*, /etc/rc.* and /etc/sysconfig/, blindly grepping for 'startx' or 'xorg'.)

What Bjørn Mork misses

Posted Dec 6, 2012 15:18 UTC (Thu) by HelloWorld (guest, #56129) [Link] (2 responses)

The problem with what Bjørn Mork says is that while it works for obviously stupid and crazy ideas, it doesn't work for ideas that are merely bad, and less obviously so.

What Ubuntu is currently doing is incidentally a good example for that. They fork off their own ecosystem, thus weakening interoperability and cooperation with other distros, which weakens the Linux ecosystem as a whole and thus ultimately themselves.

When Ubuntu first showed up in 2004, it looked as if the dream of Linux on the desktop might finally become reality. By now, it is perfectly clear to me that what they're doing to the GNU/Linux ecosystem is ultimately a disservice. It makes me feel sad and helpless.

What Bjørn Mork misses

Posted Dec 6, 2012 15:34 UTC (Thu) by raven667 (subscriber, #5198) [Link] (1 responses)

I don't know how it could have happend any other way. Canonical backed away from GNOME Shell probably because they saw what troubled direction the design was going to and couldn't get consensus to change it so fell back to Unity which re-uses some of the same technology stack but commits fewer design sins.

There was certainly a window there where GNOME2 was becoming the Linux default desktop but that seems to be all shot up now. I don't think GNOME has ever had the resources it needed, even with Canonical, to actually flesh out a stable platform and SDK that would be acceptable to J Random developer who didn't want to have to understand the implementation details of all the components that went into GNOME. There are some third-party apps which use GTK but not so many that are full GNOME apps except what ships with the OS.

What Bjørn Mork misses

Posted Dec 13, 2012 17:34 UTC (Thu) by oak (guest, #2786) [Link]

Compared to Unity, Gnome shell is actually more usable, at least on my laptops.

(In Unity, all apps menus don't fit to top bar, bar on left comes on top of windows when one tries to use elements on left side of window, window close button is on opposite side of the window than on all other OSes and distros one needs to use etc.)

Another “distribution quote of the week”

Posted Dec 9, 2012 22:28 UTC (Sun) by JanC_ (guest, #34940) [Link]

Rather, this is all further sign of a particular social problem in free software, namely the tendency to attach oneself to projects (whether that be vim vs. Emacs, GNOME vs. KDE, or systemd vs. upstart) as if they were sports teams, and then start behaving with all the public composure and mutual understanding of drunk football fans.

— Russ Allbery on debian-devel

Distribution quotes of the week

Posted Dec 13, 2012 16:59 UTC (Thu) by mirabilos (subscriber, #84359) [Link] (2 responses)

Haha, Poettering commenting about it like that… pot calling the kettle black.

Distribution quotes of the week

Posted Dec 13, 2012 20:29 UTC (Thu) by cortana (subscriber, #24596) [Link] (1 responses)

Please take such non-constructive comments to another forum.

Distribution quotes of the week

Posted Dec 14, 2012 9:57 UTC (Fri) by hummassa (guest, #307) [Link]

Well, actually... If you take in consideration the whole systemd imbroglio, he (kind of) has a well-humoured point.

Distribution quotes of the week

Posted Dec 17, 2012 16:00 UTC (Mon) by wookey (guest, #5501) [Link] (1 responses)

Maybe I should start ranting about how those terrible RedHat/Fedora people are fragmenting the ecosystem by not adopting multiarch. That would clearly be constructive. :-)

Although I do worry that as things currently stand it looks like we are heading for another Debian/Fedora (and thus derivatives) split along this axis too, which won't make life any easier for 3rd-parties. I am hopeful that other distros will come to appreciate the cleanness of the multiarch mechanism for mixed-ABI libraries and binaries, but I guess we won't know for a few years yet how ubiquitous it will become.

Personally, the more I use it the more impressed I become. It's so nice being able to _remove_ big chunks of complexity in packages and packaging that used to work around its absence.

Distribution quotes of the week

Posted Dec 17, 2012 21:22 UTC (Mon) by mathstuf (subscriber, #69389) [Link]

As a Fedora packager, I would wholeheartedly +1 a move from the current "biarch" setup to Debian's multiarch setup. I believe the ARM architecture is doing something similar to Debian here.

As for the Debian/Fedora split, there already was one since 64bit uses /usr/lib64 on Fedora and /usr/lib on Debian. So it's that they ever agreed 100% before anyways. (I do prefer /usr/lib64 and /usr/lib split for 64bit over Debian's previous "native == /usr/lib" setup which makes installing 32bit on a 64bit flavor easy, but multiarch is a better overall solution).