Richard Stallman resigns from the FSF

Posted Sep 20, 2019 18:50 UTC (Fri) by kamil (guest, #3802)
In reply to: Richard Stallman resigns from the FSF by wnowak1
Parent article: Richard Stallman resigns from the FSF

deater is complaining not so much about the OOM but about the browser being able to effectively freeze the computer for minutes at a time. I think it's perfectly fair to blame the OS for allowing something like that to happen.

deater did not imply that Linux is less stable because it's free. He was responding to Jon's apparently serious remark that "march toward world domination continues". I happen to agree that given the sad state of Linux on the desktop (arguably, we've regressed there compared to other OSes, and deater was simply providing an example of that), Jon's statement is rather preposterous.

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 19:20 UTC (Fri) by wnowak1 (subscriber, #113128) [Link] (21 responses)

When you're exhausting resources on a system, you'll get undesirable results. The fact is that the OS will kill that task and continue to run. If you want to make changes to how this works, you can adjust the memory overcommit setting to make the OOM more or less aggressive. The Linux OS has tons of parameters you can tune. There is a compromise in settings for "general use" and they are sub optimal for high end systems and low end systems alike. That is why, you the user, has the ability to make the OS perform _better_ depending on the work load and the system you're using.

How are you comparing the sad sate of the desktop? Windows or OS X require ~ 2 GB+ of RAM. I would argue that your desktop experience using an alternative OS would be worse given the same hardware constraints.

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 19:42 UTC (Fri) by Cyberax (✭ supporter ✭, #52523) [Link]

Windows 10 requires minimum 1Gb of RAM for the 32-bit version. I actually have seen it working on kiosks with about this much RAM.

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 19:47 UTC (Fri) by kamil (guest, #3802) [Link] (19 responses)

My comment about a sad state of the Linux desktop was more general, and more in response to Jon's "world domination" comment. Basically, we've lost the desktop war. And, sadly, it's not because the enemy launched an impressive attack that overwhelmed us but, instead, seemingly because we lost interest in fighting it (shouldn't the Windows 8 fiasco have been a perfect opportunity to gain some ground?).

The way I see it, desktop was always a second-class citizen in the Linux world (server workloads were always more important to, e.g., the kernel community) but now it barely even registers on the radar. Which mainstream Linux distro focuses on the desktop at this point? I don't know how far back you go but I still remember when Ubuntu first came out, and people's initial shock and disbelief at how polished its installation process and out-of-the-box desktop experience was. Back then it really felt like we were making progress in this space. And now?

Go to any Linux/free software/open source conference and count the number of people with Linux on their laptops compared to MacOS. Sadly, many of those people did go through the Linux-on-the-desktop phase, but we lost them. I know lots of people who have used Linux on their laptops for years, even decades, and are now using Macs, or even Windows 10. Now, only the diehards like the readers of LWN still bother (me included, just to make it clear).

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 20:18 UTC (Fri) by pizza (subscriber, #46) [Link] (1 responses)

> Go to any Linux/free software/open source conference and count the number of people with Linux on their laptops compared to MacOS.

It's a reflection of corporate IT policies more than anything else.

Using a Mac allows those users full access to corporate IT infrastructure (email, VPNs, support, etc), the hardware is (or at least used to be) generally sane, and they can run Linux in full-screen VMs if need be. As an added bonus, they don't ever have to fight driver or hardware issues.

One other thing -- don't forget that most for most "open source" developers, Linux is just an implementation detail hidden beneath three separate frameworks and a devops container system that SomeoneElse set up.

Meanwhile they are probably using a "code editor" that's actually a javascript application running in a cut-down web browser, that has no real interaction with the local operating system beyond the method used to launch it. Even the source code (and revision control) is stored/managed in the cloud.

....Feeling old yet?

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 22:06 UTC (Fri) by kamil (guest, #3802) [Link]

> It's a reflection of corporate IT policies more than anything else.

That may be your experience but it isn't mine. I work in computer science research; there are few restrictions in such places when it comes to what hardware or OS you can use (IT people have learnt over the years to just "let them play with whatever they want").

Virtually every researcher I talked to who switched away from a Linux desktop said that it was because they got sick and tired of iffy hardware support, bugs that never get fixed, and missing basic features that everybody else takes for granted. To them, Linux desktop was basically abandonware. Sure, other factors also played a role, such as a need to run MS Office (you would be surprised how many CS people don't like LaTeX), although these days they could of course do that in a VM.

I certainly agree with you though that a lot of development these days takes place so far above the kernel that the OS simply doesn't matter...

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 20:22 UTC (Fri) by rahulsundaram (subscriber, #21946) [Link]

Long before Ubuntu, Red Hat Linux, Caldera, Corel, Mandriva etc tried the consumer desktop thing at one point or the other and they didn't continue for the same reason, Canonical eventually pulled the plug - there doesn't seem to be a viable business model around it. Red Hat continues to invest significantly in GNOME but is certainly not selling a consumer desktop edition anymore.

The only major success related to space is Chromebooks. Hardly a traditional desktop. Browsers seem to winning the desktop war to the extend there is even a desktop market at all and many of the desktop clients are thin wrappers around web based tech for what it is worth.

Lots of folks have moved on to using mobile phones and some of them are using tablets where they would have been using laptops before. So that's part of what happened as well.

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 20:27 UTC (Fri) by wnowak1 (subscriber, #113128) [Link] (12 responses)

According to itsfoss, Linux has the largest desktop market share ever, still third but it's much better than FreeBSD for example.

As far as who / which distro focuses on the desktop, there are several distros that aim to make a pretty and usable desktop.
https://fossbytes.com/most-beautiful-linux-distros/

Chromebooks/ChromeOS is another one if you want to count that.

I don't disagree with you that desktop was/is a second-class citizen but the point is that to ding Linux because a userspace application is killed while running out of memory on a low end system is a bit unfair. It actually doesn't even matter that it was Firefox. It could be anything, apache, or some other software running out of memory and being terminated by the oom killer. I rest my case ...

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 21:02 UTC (Fri) by deater (subscriber, #11746) [Link] (11 responses)

> I don't disagree with you that desktop was/is a second-class
> citizen but the point is that to ding Linux because a
> userspace application is killed while running out of memory
> on a low end system is a bit unfair.

You are misunderstanding. I *wish* the OOM killer would kick in and kill firefox. What happens on my machine is that once the system hits OOM conditions, the desktop soft-locks, sometimes for over 30 minutes, with the system unresponsive.
Sometimes if I hit control-alt-f1 fast enough it will eventually (maybe after 5 minutes) switch to a console window where I can kill firefox manually if I've left a logged in root console there.

This is not disk thrashing, I have swap turned off as I have an SSD drive.

Linux is completely failing in this case, and it's bad enough that after 23 years of Desktop Linux use I'm considering switching to something else.

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 21:12 UTC (Fri) by pizza (subscriber, #46) [Link]

> This is not disk thrashing, I have swap turned off as I have an SSD drive.

Well, that's heavily exacerbating your problem...

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 21:17 UTC (Fri) by kamil (guest, #3802) [Link] (1 responses)

You are wasting your time. He will try to tell you, as he already did, that there are all those /proc or /sys files you can tweak to make OOM behave better and that it's your fault that Linux is not optimized for your low-end hardware.

As I see it, Linux is not optimized for any desktop hardware. It's optimized for servers. In fact, over the years, it has become more and more desktop-hostile, to mention anti-features pushed on us by enterprise distros such as the "predictable" network interface names as one example. That is why people who used it on desktop for years are switching away -- because they don't see any future in it, any light at the end of the tunnel.

But there's no convincing some people that there is a problem.

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 23:56 UTC (Fri) by HelloWorld (guest, #56129) [Link]

Why would any desktop user care about network interface names? You use NetworkManager and you're done, and it has been that way for ages now.

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 21:29 UTC (Fri) by wnowak1 (subscriber, #113128) [Link] (1 responses)

So you've invested in an SSD, but still stuck to 4 GB of RAM for your everyday desktop use?

You should consider enabling swap.

"A swap file is space on a hard disk used as the virtual memory extension of a computer's real memory (RAM). Having a swap file allows your computer's operating system to pretend that you have more RAM than you actually do."

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 21:31 UTC (Fri) by wnowak1 (subscriber, #113128) [Link]

You can turn off swap if you have lots of RAM. The opposite is true if you have little RAM.

Richard Stallman resigns from the FSF

Posted Sep 23, 2019 19:56 UTC (Mon) by flussence (guest, #85566) [Link] (5 responses)

You may not have swap *allocated*, but you're absolutely still swap thrashing. The only difference is that the system's reverting to PDP-11 era behaviour of swapping in and out hot program code from /usr/bin with every context switch because it has nowhere to flush cold data.

And that's *far* more likely to kill your SSD.

Richard Stallman resigns from the FSF

Posted Sep 23, 2019 20:03 UTC (Mon) by kamil (guest, #3802) [Link] (4 responses)

That can't be correct. Without a swap configured, there's nowhere to swap out to.

Yes, clean pages will be dropped more often, requiring them to be read again from files in /usr when they are needed, as you said, but nothing will be written there, so his SSD will be just fine.

Richard Stallman resigns from the FSF

Posted Sep 23, 2019 20:34 UTC (Mon) by farnz (subscriber, #17727) [Link] (3 responses)

Most pages of demand-paged executables are clean. Without swap configured, all the system can do is discard random pages of an executable (or any other file-backed mmap area) in the knowledge that it can page those back in from the source file as required.

All you do by removing swap is force the system to page out file-backed mmap pages such as executable code in preference to anonymous pages; this is often not what you really want, as paging out a small amount of anonymous data that's not been used in a while can be enough to permit the system to exit paging thrash. See Chris Down's essay in defence of swap for more details on why swap is needed.

SSDs, erasing, and wear rates.

Posted Sep 24, 2019 9:12 UTC (Tue) by james (subscriber, #1325) [Link] (2 responses)

What you write is correct (to the best of my knowledge).

However, it is erases that wear out an SSD. If you are configuring your system solely to minimise SSD wear, many reads from /usr are far less harmful than one write to swap.

Of course, if you are optimising a system solely to minimise SSD wear, why not get rid of the SSD altogether and go back to spinning rust? We normally buy SSD-based systems for performance reasons; it seems inconsistent to then not take performance into account when configuring it.

SSDs, erasing, and wear rates.

Posted Sep 24, 2019 12:44 UTC (Tue) by deater (subscriber, #11746) [Link] (1 responses)

*sigh* I should have abandoned this thread a while ago.

anyway, thanks everyone for quoting the wikipedia page on swap files to me, believe it or not I know what they are, how they work, and I have even written my own VM-enabled custom operating system before.

> why not get rid of the SSD altogether and go back to spinning rust?

Sure, next time I rip open this 5-year old macbook air for maintainence maybe I'll shove in some huge 5400rpm 3 1/2" drive. Maybe I'll also solder in some DIMM slots so people can stop accusing me of not having enough RAM.

Really, if your operating system is so poorly written it can't run in 4GB on a multi-gigahertz machine, maybe you need to step back and re-evaluate your coding a bit.

SSDs, erasing, and wear rates.

Posted Sep 24, 2019 13:21 UTC (Tue) by pizza (subscriber, #46) [Link]

> anyway, thanks everyone for quoting the wikipedia page on swap files to me, believe it or not I know what they are, how they work, and I have even written my own VM-enabled custom operating system before.

...In other words, you deliberately and knowingly chose to mis-configure a system designed around overcommit of memory in a way that causes it to break on your particular application workload?

> Really, if your operating system is so poorly written it can't run in 4GB on a multi-gigahertz machine, maybe you need to step back and re-evaluate your coding a bit.

Just FYI, passive-aggressive insults are not the way to get folks to care about your self-made predicament. You may want to step back and re-evaluate your approach.

Richard Stallman resigns from the FSF

Posted Sep 20, 2019 22:55 UTC (Fri) by da4089 (subscriber, #1195) [Link] (2 responses)

I think the "problem" with Linux on the desktop is one of confidence.

On servers, on phones, on embedded devices, developers have made Linux the best solution available. This has required both application and kernel work to produce a product that is better than the alternatives.

On the desktop, I think Linux is hamstrung by its limited vision. We seem to always aim to be "as good as" Windows or macOS. This ends up dooming the effort to be, at absolute best, a clone of the last release of those environments.

Desktop Linux needs to envisage a workspace *beyond* what its competitors provide, and deliver it. If that vision is compelling, users will switch. If not, why would they? Users migrate from iOS to Android, from Windows to macOS and macOS to Windows -- different applications, *incompatible* applications aren't an issue, it's the perception that the end goal is better that's required.

Richard Stallman resigns from the FSF

Posted Sep 22, 2019 10:12 UTC (Sun) by anselm (subscriber, #2796) [Link]

Desktop Linux needs to envisage a workspace *beyond* what its competitors provide, and deliver it. If that vision is compelling, users will switch. If not, why would they? Users migrate from iOS to Android, from Windows to macOS and macOS to Windows -- different applications, *incompatible* applications aren't an issue, it's the perception that the end goal is better that's required.

People do move between iOS and Android, or between Windows and macOS, but they do that by buying new hardware that has the other operating system already on it (in fact you can't even install iOS on hardware that is intended for Android and vice-versa etc.). That is very difficult in the case of desktop Linux as there are generally no computers with pre-installed Linux available where most people go to buy a new computer. Every computer in these places, however, has a reasonably adequate operating system pre-installed already, and Linux would have to be very compelling indeed in order to get people to go to the trouble of installing it themselves, voiding the warranty, and so on. Finally, whether people switch between iOS and Android, or Windows to macOS, exclusively because the operating system's “vision” is so “compelling” is by no means clear – for many people it's a question of what's being offered at what price, whether the hardware looks sleek and well-designed, the availability of application software and peripherals, and (certainly in the case of Apple) a lifestyle choice just as much as a technology choice.

In my experience, people have no problem using desktop Linux if someone knowledgeable installs it for them. In fact, most of the desktop users I support tend to prefer it to Windows once they've got the hang of it – KDE, for example, works pretty well and does offer useful features that Windows doesn't. So as far as I'm concerned the problem isn't “confidence”, it's putting actual Linux desktops in the hands of actual users.

Richard Stallman resigns from the FSF

Posted Sep 23, 2019 15:09 UTC (Mon) by raven667 (subscriber, #5198) [Link]

> On servers, on phones, on embedded devices, developers have made Linux the best solution available. This has required both application and kernel work to produce a product that is better than the alternatives.

That work didn't just happen, it was paid for (by volunteers and businesses) to the tune of hundreds of thousands of full-time workers over the years, the sophistication you see is the direct result of the amount of effort put into it. The desktop, while full of passionate developers and containing some businesses, has maybe only had thousands of full-time workers so the scope of the vision needs to be constrained by the reality of the resources available, and comparisons to Apple, Microsoft or Google need to be understood on the relative efficiency of the development effort rather than the absolute amount of work that gets done. Its absurd to complain that the work of 100-500 (GNOME, KDE, XFCE, etc.) developers is less sophisticated than the work of 5,000-10,000 (Apple, MS, Google), but the fact that they can be compared at all is an achievement.

> Desktop Linux needs to envisage a workspace *beyond* what its competitors provide, and deliver it.

I would say that ChromeOS is probably the closest to a success story here, but part of the reason why is that the desktop matters less than it has in the last 20+ years and most current end-user development is in cross-platform JavaScript web browser applications and not in native desktop ones.