The plan for Linux after Torvalds has a kernel of truth: There isn’t one

The Linux kernel is a remarkable creation. It has achieved a fundamental status in the industry, and thus the world, unmatched in scope, stability, and reputation. It powers lightbulbs to supercomputers, not to mention the billion-plus global army of Android. It covers a host of processors, a massive array of supported devices …

[jake]

If Linus goes TITSUP[0] tomorrow, Linux will carry on.

It's all been discussed ad nauseam, starting a couple decades ago.

Look up "what happens if Linus gets hit by a bus" for just a glimpse of the total conversation, and roughly what will happen.

tl;dr version: There will be an up-to-date and modern mainstream FOSS Linux kernel available that is not under the control of any corporate entity for far longer than you or I will be around to worry about it.

_{[0] Torvalds Inconveniently Totally Stops User Processes}

[A Non e-mouse]

The article is a bit click-baity.

Although Linus is the big-cheese for the Linux kernel there is a whole army of people working on the project. "All" that Linux does is act as overseer. Others are ready to step up if Linus logs off.

[jake]

Of course it's click-baity.

ElReg is a red-top, and it's the silly season. They are required to post such articles or they will get thrown out of the guild.

_TING

[Yet Another Anonymous coward]

>"All" that Linux does is act as overseer.

That's the important job.

Yes there are a whole army of people working at Redhat that will support Redhat, and at AWS and Azure and etc

The tricky bit is ensuring that we don't have a Redhat Kernel with Redhat fixes theoretically available to anyone who writes to Redhat, and an Azure kernel and an AWS kernel and a dozen incompatible "People's front of Judea" and "Judean People's Front" kernels

[Anonymous Coward]

Isn't systemd our new overseer ?

[Yet Another Anonymous coward]

Don;t worry systemd will soon be absorbed into Emacs

[keithpeter]

Just had an image of RMS and Agent P discussing some arcane issue on emacs-devel.

See icon.

[jake]

There is no RedHat anymore. It's IBM.

And IBM already releases proprietary IBM Linux kernels branded RedHat. This doesn't affect the mainline kernel in any real way.

[Yet Another Anonymous coward]

There is no IBM anymore, it's just the unprofitable part of Redhat

[CountCadaver]

Should just have sold the entire company to Lenovo when it still had some sort of value....?

[CountCadaver]

A bit like Yugoslavia was held together due to Tito managing tensions via various means, the problem was no defined succession and the whole lot violently came to an end when long simmering tensions exploded.....

[Excused Boots]

Yes you are right, nobody is claiming that Linus does all off the work and coding himself, but in the event of conflicts, then the Linux community does tend to defer to what Linus wants. Fine.

But assume that he falls under a bus, decides tomorrow that he can't be bothered anymore (both unlikely but not completely impossible), who does ‘step up’? And whoever they are will the general community respect his or her input as much as they do for Linus?

Honestly, I hope I’m wrong but I do genuinely suspect that the entire Linux system will fracture with various versions ever so slowly moving away from each other. It will take time but imagine a future where one version of Linux won’t talk to or is not code complaint with another version!

Both will claim to be the ‘true inheritors of Linus’s intentions’.

Now I know this will be massively downvoted, so be it. But before you do click the downvote button, just stop and think! Honestly do ask yourself what you think will happen in the above eventuality?

[Anonymous Coward]

Honestly, I hope I’m wrong but I do genuinely suspect that the entire Linux system will fracture with various versions ever so slowly moving away from each other. It will take time but imagine a future where one version of Linux won’t talk to or is not code complaint with another version!

But what's stopping that happening already? Presumably, the fact it hasn't means it ultimately wouldn't be beneficial to anyone, regardless of whether Linus still has a heartbeat.

One downside that immediately springs to mind is that a fractured linux ecosystem would lead to a fractured talent pool which would make it difficult for corporations to poach star developers from their competitors.

[A.P. Veening]

But what's stopping that happening already?

Not what but who: Linus Torvalds.

[Zolko]

We can see the cracks coming already : systemD, Rust in the kernel, user-space drivers ... once Linus is gone, who will decide if and how Rust should be part of the kernel ? Or how much the kernel should rely on systemD being there ?

[Missing Semicolon]

The power of Grump

The unique feature of Linus is that if someone produces rubbish, or someone proposes something that makes stuff worse, he can just go "No! Because I Say!" and he has the respect of the community to be heard. No-one else has, or will, earn the kind of respect he has. So Linux will turn into the usual FOSS mush, with bits going proprietary. a leadership group more concerned with appearances than progress or quality, and the humble contributor ignored.

Who would replace him? My guess would be a placeman from IBM or Microsoft. Agent P? Not beyond the bounds of possibility.

[A.P. Veening]

Re: The power of Grump

Who would replace him?

I'd say a (small) committee of respected experts, preferably with an odd number of members, working preferably by consensus. A nice example (in a totally unrelated field but with equally massive egos) is the ECB, especially the first period under Wim Duisenberg. In the beginning he flabbergasted the whole economic world by striving for consensus, after a while they appreciated the results.

[A.P. Veening]

We can see the cracks coming already : systemD, Rust in the kernel, user-space drivers ... once Linus is gone, who will decide if and how Rust should be part of the kernel ? Or how much the kernel should rely on systemD being there ?

There already is Rust in the kernel (with Linus' blessing), so that isn't a discussion anymore. There are also some user-space drivers that haven't been approved to/for kernel-space (yet). As for systemD, Linux is already split on that and several branches work perfectly fine without it, so the kernel shouldn't rely on it.

[Uh, Mike]

Guess what:

Somebody owns the Linux trademark.

Somebody will inherit it.

[jake]

"There already is Rust in the kernel (with Linus' blessing), so that isn't a discussion anymore."

Kinda. In reality, what little rust is in the kernel is, at most, a statistical aberration, and even that is aggrandizing it. And the "blessing", as full of "we might" and "maybe we will" and "perhaps" and "eventually", and "drivers, probably" etc. etc. as it is sounds more like throwing the braying fanbois a bone just to shut them up. Note that nowhere does Linus say "Let's do it" or "We are going to" or "It will be soon".

As for the init ... THINK about it. During boot, the kernel sets up for the hardware it has available to it ... and then, and only then, does it pass control to the init. There is no need now, and never will be a need for the kernel to rely on any one specific init. Linux (the kernel) is NOT split on this. Some ill-conceived Linux (+GNU) Distributions might be, however.

[MacroRodent]

Who

> Not what but who: Linus Torvalds.

The license of Linux permits anyone to start an own fork and make releases. However, good luck getting anyone to contribute only to it, unless it server some specialty that the Torvalds Linux does not. For example, the Android kernel was that way for a long time, but it has gradually merged most (all? not sure) of its special sauce to the master, to simplify maintenance.

In a way, splintering is prevented by network effects, not Linus.

[Anonymous Coward]

Not what but who: Linus Torvalds.

No, you're wrong. Linus can't stop someone or, as posited by the OP, a corporation, from spinning their own kernel variant.

[A.P. Veening]

No, you're wrong. Linus can't stop someone or, as posited by the OP, a corporation, from spinning their own kernel variant.

While you are technically correct, the one serious variant there was (Android) returned to the fold. And that was partly because Linus just kept the whole rest together.

[Captain Zippy]

...which is _exactly_ the case _right_ _now_.

Death of the internet^H^H^H^Hlinux predicted. Film at 11.

[Captain Zippy]

There's are already fractured version of 'linux' - not necessarily all at the kernel level, but this is what distro's are. We can cope with them, and we will vote with our feet and keyboards. Crap stuff will die off, and good stuff will get adopted.

The GPL means that we will all (generally, handwavy) benefit from these alternate branches.

[CountCadaver]

Though civil wars can and have erupted over lower tensions than those within software communities.....

[Anonymous Coward]

It might not be under the control of a single corporate entity, but Linux fell under corporate control long ago. It does exist only because a lot of corporate entities need a low cost kernel and OS, and thereby fund its development to avoid to pay far more expensive licenses or direct investments to build their own.

But keep on thinking it is developed by monks in a mountain monastery, for the benefit of the world...

[Doctor Syntax]

That makes a nice change from you telling us it's all put together by amateurs.

If you can just get over the notion implicit in the last line that that's what the rest of us were thinking...

[Anonymous Coward]

Never said the Linux kernel is made by amateurs. I wrote it has an outdated and lame design still stuck in the 1970s, but it's not written by amateurs. Those corporates running it don't want to have a system written by amateurs to run their server applications on - that's how they make money, after all.

The Linux desktop GUIs are made by amateurs, and most of its applications (not the pro, paid for ones, of course) - since the same corporates don't really need a desktop system and desktop applications, which would hide precioussss data from them, everything must happen online to hoard data better. So let amateurs fil that niche, and keep it a subpar experience. Even Nadella is killing Windows and Office to force people to work more online....

Look beyond your nose sometimes, you may discover that things look a little different.

[ovation1357]

The current GNOME desktop certainly isn't made by amateurs - it's mainly controlled by professionals employed by RedHat/IBM... It's also utterly terrible and feels like it's been created by total amateurs. Interesting....

It's not really fair to point at Linux GUIs and Desktop Environments when you've got the "professionally" designed and developed clusterf**k of Windows 8, 10 and 11 and Office applications. Awful, awful GUI (although not quite as awful as GNOME) - flat, low contrast, ugly mess which can barely be customised and often has oversized text and/or masses of wasted whitespace, and even a smattering of amateur errors such as text alignment problems.

[Paul Hovnanian]

"It's also utterly terrible and feels like it's been created by total amateurs. Interesting...."

The Arc was skippered by amateurs. The Titanic, by professionals.

[Nick Ryan]

In the defence of the Titanic: it was at least real.

[Anonymous Coward]

Arc? Why would an arc need a skipper? Unless you mean the alleged Ark.

[Anonymous Coward]

I exactly wrote Nadella is trying to kill Windows desktop. Windows UI has been dumbed down to Linux levels, and became far less capable and friendly. That's undeniable. And the reason? Cut costs putting it in the hands of less skilled developers, eliminate testers. MS too no longer wish to make money from desktop OS and desktop applications. Everything must move to Azure, and thereby the investments are there (and AI of course). Just like the companies controlling Linux are doing in the past twenty, twenty-five years. Huge profitfs for them (and some of their developers), worse applications for users (and some free labor exploited).

BTW: Windows 8 was an excellent design for tablet and phones. Those who designed it knew what they were doing. It should have not ever, ever, ever been used on a desktop system.

And I hope GNOME developers are interns, because if they are "professional" GUI developers, they should look for a totally different job. Truly a manual about how a GUI should NOT be designed. But again, where RedHat/IBM make money? Desktops, or server/services?

[Anonymous Coward]

You have never used Linux, have you? It's been some Linux DEs that have moved towards Windows (my default is always LXDE because it doesn't get in my way and works sensibly over rdp (yes, I'm a perv, I use Linux over rdp on Windows :D)), not the other way around, and both are probably trying to become a mobile interface.

[Paul Hovnanian]

"Windows UI has been dumbed down to Linux levels,"

Keep making wisecracks like that and we're taking your GUI away and replacing it with a command line.

[Captain Zippy]

even worse - a CMD line.

[ovation1357]

"Windows UI has been dumbed down to Linux levels,"

s/Linux/GNOME/g

Then maybe I'll agree, but using the word Linux isn't just wrong; it's fighting talk!

[Grogan]

Why is it a lame outdated design, because it isn't a "microkernel"? Does Polly want a cracker? (posturing parrot)

[chololennon]

Don't waste your time

> Why is it a lame outdated design, because it isn't a "microkernel"?...

Don't waste your time, he/she is a well known Linux hater, just check his/her posts in this site.

[Mage]

Re: Microkernel

Andrew Tannenbaum argued that Torvalds had made a mistake. I played with his Minix back in the day, which is still around.

https://en.wikipedia.org/wiki/Tanenbaum%E2%80%93Torvalds_debate

I once was in the Tannenbaum camp, and possibly for a real-time embedded system in HW that never gets an update, the microkernel might be better.

If NT (Version 3.1 came out about the same time as the Linux Kernel alone) was ever microkernel, it wasn't by NT 4.0. Before NT 4.0 a BSOD was rare, not because NT 3.51 was microkernel (it probably wasn't) but with NT 4.0 they moved graphics driver (and thus printing) into the Kernel. NT 4.0 with curated graphics and printer drivers didn't have BSODs, at least I don't remember any.

It's not a simple issue and it's not a simple either/or. There are issues to do with CPU type/architecture, driver performance and RAM. Note a 4 MHz 8 bit CPU could be running a RTOS and able to do RC5 remote control decoding or serial I/O in software and Linux, MacOS and Windows unable to do it reliably without a clever device driver as the standard versions of those are not RTOSes, no matter how fast the CPU is. There is actually a RTOS version of Linux.

The security issues and stability are not so black and white either.

In a sense BSD and GNU wouldn't exist except for AT&T / Bell labs claiming complete copyright, when in fact a lot of it wasn't funded by them. Linux is just the Kernel, so a Linux distro in a sense 1970s. That doesn't make it obsolete. Concrete was invented by the Romans and fired clay (bricks, pottery) is older. Neither is obsolete.

Examples

https://en.wikipedia.org/wiki/Microkernel#Examples

Includes Mac OS7 to Mac OS9, all of which were poorer at multitasking and security than MacOS X (based on NextStep, based on BSD)

Symbian is microkernel but the issues writing applications was the fact it was only using the Mobile Java licence (unlike Android which effectively used Desktop Java) and the Nokia GUI went backwards after 2003 with S60 and the scrapping of S80 and other developments. The success of Android (using modified Linux Kernel) and failure of Symbian is nothing to do with kernel architecture, but Nokia and Google political decisions.

ANYWAY

It's all irrelevant to the future of Linux after Torvalds. It's survived the Microsoft Halloween papers (1998?) and "Linux is a cancer" claims. Windows was over 90% of servers on Internet once. Linux is now 98%. MS had nearly 50% of PDA/phones in USA once and now is nowhere. Home computing was once dominated by MS, now tablets and phones, that don't run Windows.

Many setboxes ran Windows CE. BT swapped theirs to Linux via OTA updates.

TVs, projectors, home routers, ereaders and almost all tablets and phones not made by Apple run either actual Linux, or Android with the modified Linux Kernel.

[Mage]

Amateurs vs Professionals

The main difference sometimes between an amateur and professional isn't expertise, or attention to detail (avoiding bugs), or proper documentation, but can simply be pay. There are plenty of huge corporations doing incompetent things. Or evil things. Profit and empire building rather than excellence is the motivation.

The original distinction between Amateurs vs Professionals was pay.

[Paul Hovnanian]

Re: Amateurs vs Professionals

"The original distinction between Amateurs vs Professionals was pay."

That's the definition. But how does that relate to the difference in performance between these two classes? If not paying you results in your not sinking my boat, I'm OK with that.

[mirachu]

You ascribe to professional what often isn't there (competence for one). Also, many of the contributors ARE those professionals.

Professional only means you get paid, nothing more.

[O'Reg Inalsin]

Explicit

[Anonymous Coward]

I guess the concern is less to do with whether someone can take the job, but rather whether they'll have the... let's say fortitude to keep the project going in the direction that it has been. We've seen many once promising organisations bend to convenience and pressure to perform.

So the question is, will the next Benevolent Dictator be able to tell an Nvidia or Google engineer to fuck off with their sloppily written/conceived code?

[m4r35n357]

Just a guess; whoever takes over (Poettering?) will be forced to be visibly "nice" for appearances sake (as if Linus isn't one of the nicest blokes there is!), and quality will plummet.

[JoeCool]

Is Poettering a Kernel developer ?

[m4r35n357]

Not yet

[jake]

Was. Got kicked out for not playing well with others.

[ThatOne]

> whether they'll have the... let's say fortitude

Indeed, that's what I thought too: Many kernel devs surely have the required technical qualifications, but his successor would first and foremost also need the "special character" (or temper...) required to be able to herd a large gaggle of developers, keeping them focused and preventing mission creep, personal vanity projects, or commercial interests taking over. I think it's the biggest and most difficult part of his job: Having a vision, and the chops to pull it through.

[Phil O'Sophical]

Almost sounds like a proprietary product...

[ThatOne]

Well, imagine a ship with a dozen helms and as many helmsmen. Each with his/her personal idea of when the ship should head to. Will it be able to get anywhere? I don't think so.

Fortunately OSS isn't like that. There is a boss (maintainer) who says (rightly or wrongly) where the project is heading to, in other cases there are so few people working on it that it doesn't matter. The freedom of OSS is not in that everybody is allowed his opinion, it's that if you aren't agreeing with the current official boss, you are free to copy the source code and try to do it your own way.

[GNU Enjoyer]

You joke, but Linux was originally a proprietary product in 1991 and resumed being one in 1996 with the inclusion of the first proprietary program (of many).

Linux today contains a lot of object code without source code (or patches to object code without source code) disguised as arrays of numbers, undocumented init sequences and many drivers where ½ the driver is in linux.git and the other ½ is only available as a proprietary binaries from "linux-firmware.git"

[Captain Zippy]

Thing is, AFAIK. - Linus doesn't control developers. They can do anything they want. What he does do, is control which submissions make it into the mainline kernel - for reasons of quality and code coherence (mainly)

Which is not the same thing at all.

[vogon00]

Now, I don't know jack about succession - apart from the fact it's a risky time.

The sole reason that the Kernel - and all that depends on it - has been so so successful is that there has *been* a Benevolent Dictator. The fact that that is currently the guy who started this Linux kernel thing is incidental (Well, it's not, but you know what I mean!).

Whenever Linus logs off, there needs to be someone with the same level of independence, mission-awareness, intestinal fortitude and altruism as Linus to ensure that Kernel development take place with the goal of benefiting the entire ecosystem, not just a greedy corporate sub-set.

In particular, whoever takes over must insure that the Kernel remains build-able from source by anyone, i.e. completely without BLOBs as far as possible.

Linux works form many reasons, but mainly because of the ethos behind it. IMHO, Linus' replacement *must* continue to act as a dictator...preferably a benevolent one. Change that idea, and you will change Linux beyond recognition...it will end up as the same corporately-driven and competitive fuck-up that is the rest of the software world. The trick is to find someone who has enough smarts to keep development both flowing and neutral to ensure a level playing-field, for the benefit of all.

Personally, I'd be starting the search NOW, and handing over sooner than later to ensure that the new body has the time to become accepted in the role by the rest of the ecosystem.

[Anonymous Coward]

Benevolent dictator = SPoF

Just wait...

[hoola]

I supposed there is a considerable risk that corporate money will win and some existing (or new VC backed) outfit will try and take control.

I know everyone will hold their hands up and say "impossible" but that is not the case. We have seen time and time again that once there is a sniff of an opening corporate money men will be in there like a rat up a drainpipe.

One could very easily argue that the Linux Kernel is one of the most valuable assets that exist............

That means ££££££££ or more accurately $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$

[GNU Enjoyer]

I really doubt that is the case

As most of the reason for Linux's functionality and popularity seems to be how it was built in top of GNU (as most of the work was already done) and how over a thousand developers have volunteered (which is extremely rare).

It certainly wasn't solely because of one person - even if that person is given credit for absolutely everything (most of the credit he receives is for things he didn't do and usually isn't anything to do with Linux either (but he accepts it anyway)).

Eventually Linux will likely die, whether Linus steps down or not, due to all the proprietary software and license infringement (that is left to fester) and how it is (partially) GPLv2-only - such kind of things kill a project eventually (there's a reason you don't see GPLv1-only projects anymore - you can fix license bugs or you die).

Although the chances are slim, Linux could die next week if some court next week made an error that invalidates the GPLv2 (as the -or-later escape raft has been punctured 1800 times).

>all that depends on it

Most software does not in fact depend on Linux - most packages in Debian run fine (just a little slower) if you replace Linux with Hurd.

>whoever takes over must insure that the Kernel remains build-able from source by anyone, i.e. completely without BLOBs as far as possible.

Currently already Linux contains proprietary software (object code and patches to it without source code, undocumented init sequences) and many drivers do not function without proprietary software.

The only project that has ensured Linux is build-able completely from source is GNU Linux-libre; https://gnu.org/software/linux-libre

https://www.fsfla.org/ikiwiki/selibre/linux-libre/wash.png

Such project will be continued for as long as Linux remains the primary kernel for GNU - if Hurd or another free kernel supplants Linux, such package will be eventually decommissioned.

[wolfetone]

TITSUP?

Don't you mean: Total Inability To Speak Usual Profanities?

[Anonymous Coward]

I seem to remember a linux mag doing a chain of sucession

Maybe Linux Format or something of that nature.

[DS999]

Of course it will continue

The question the article is REALLY asking is, will just one "Linux" project continue after Linus is gone?

Or will there be disagreement on who takes over, or even if there was agreement as far as "who" perhaps not long after disagreement over the decisions he/she is making. Being able to fork a open source project is an advantage when it comes to working around damage like a corporate takeover of the development process or the person/people responsible for issuing new releases or posting them in an official location quit doing so for whatever reason.

But it is also a disadvantage in that everyone with a grudge over the way things are being done can go their own way. If it is just one malcontent who overestimates how many people agree with him, his fork will gain little traction and isn't cause for concern. But what happens if there's a big rift and it splits things roughly in half - half of the driver maintainers go one way or another? Things would get worse for both projects at least in the short term until they found new maintainers to fill their holes, and would probably continue to suffer somewhat into the long term as the best and most knowledgable people are now split between projects.

If Linus merely retires I don't see this as an issue - he might give up leadership and a day to day role but it isn't like he's not going to be involved at all. But if he leaves due to death or serious illness, I see a real chance that a rift could eventually develop.

[Anonymous Coward]

You've never been in a social club have you?

When there's one person, a truly benevelant leader, then they keep the factions and hot-heads in line. When that leader leaves the in-fighting begins, everything stalls and arguments escalate. I will put money down that when Linus has had enough it will be a shit-show for a while, a few weeks, months or even a couple of years until someone or some org steps in to bring back order.

[Persona]

Linux will carry on. Your nightmare scenario should be that there will be two or even worse "many" up-to-date and modern mainstream FOSS Linux kernels available that are not under the control of any corporate entity.

[A Non e-mouse]

built by volunteers

Linux started out as a volunteer project but it's now a collaborative corporate affair. If you look at the statistics, most Linux kernel work is by people paid to work on it. See the development stats for the last kernel release: lwn.net/Articles/1031161/

[Doctor Syntax]

When I got to that line I knew I didn't need to read any further.

[jake]

They are paid to work on it by $CORPORATE_ENTITY, yes. But that doesn't mean they are not volunteers.

I've been on the LKML since the year dot, and I do not remember anyone griping that they were only there because their Boss forced them into it.

[Alistair]

I've been on the LKML since the year dot, and I do not remember anyone griping that they were only there because their Boss forced them into it.

Again proof of selective memory issues on certain issues.

[jake]

Please, refresh my memory Alistair.

Where exactly can I find people on the LKML bitching about their Boss forcing them into kernel coding, despite them not wanting to be doing that?

_{They say the mind is the first thing to go, and I'm not getting any younger ...}

[Anonymous Coward]

Microkernel

After all these years I still lament the choice of a monolithic over a microkernel architecture for Linux. Even though Linux has shown the enormous latent demand for a free and open-source operating system we're still being denied many of the advantages of micokernels, like security and stability.

The stability of Linux is the result of may hours of hard work by many of the individuals involved, but it could easily become a nightmare if there's one misbehaving driver or component since everything runs in the same address space. This is especially crucial for embedded systems which sometimes need to run for decades and never get updated. In those cases it would be much more prudent to use an architecture which is more or less secure by design.

[A Non e-mouse]

Re: Microkernel

If you look at the way BPF is developing, you could argue Linux is heading in that direction.

[Dan 55]

Re: Microkernel

On the other hand if Linux were a microkernel and had a stable ABI for drivers, there wouldn't have been as much incentive to open source the drivers. Many more drivers would be like Nvidia's binary blobs shoehorned into the kernel, but modular so they wouldn't even have to be supplied as part of the kernel.

[Anonymous Coward]

Re: Microkernel

I see your point, but companies will always make an effort to keep their trade secrets a well..secret.

And there will always be individuals who'll attempt to make an open-source driver. There are open-source drivers for NVIDIA but I'm unsure if anyone's using them since they lack performance and features compared to the closed-source one.

[JulieM]

Re: Microkernel

No. Microkernels are an abstraction too far.

Processes in user space that want to interact with hardware can only interact with an idealised version thereof, provided by driver code within the kernel, precisely so that hardware cannot be interfered with directly by anything happening in user space.

The proper place to put a fence is where as little as possible has to pass through it, not just where it looks pretty.

Now, maybe there's an argument to be made for an extra layer between user space and kernel space, just for driver code; but drivers absolutely do not, under any circumstances, belong in user space.

[Yet Another Anonymous coward]

Re: Microkernel

>No. Microkernels are an abstraction too far.

Microkernels don't go far enough, you need KAAS

Kernel as a service allows your OS to rent a new kernel from a marketplace of online suppliers for each new process

[Anonymous Coward]

Re: Microkernel

QNX begs to differ.

As do MURTL, Minix and many others.

I'm currently reading up on QNX as I find it fascinating. I believe an older version was made open-source only for the license holders to change it back to closed-source licensing again.

[Jamie Jones]

Re: Microkernel

There used to be a QNX desktop distribution delivered on a single 1.44MB floppy.

You can still get that https://archive.org/details/QNX_incredible_1.44m_demo_v4.0

[Anonymous Coward]

Re: Microkernel

I can't get it to grab the mouse. It's pretty damn fast and responsive even running under an emulator made in WASM in a web browser.

[Anonymous Coward]

Re: Microkernel

I don't understand your argument since every operating system virtualizes hardware through an API. The only difference is that in a microkernel OS the hardware API is exposed through a messaging interface.

[thames]

Re: Microkernel

The Free Software Foundation (GNU) said the same thing as you about microkernels being the way forward. After they developed all the other bits of the operating system, from shell to utilities to file system, to compilers, they set about work on the final bit, the kernel which was to be a microkernel.

While everyone was waiting for that, a university student in Finland posted on a mailing list that he had developed his own monolithic kernel which may be not quite as professional as the FSF's effort, but was available now if anyone was interested in his modest efforts.

And this is why today we have GNU/Linux, with the kernel from Linux and everything else from GNU.

Meanwhile the FSF are still working on a microkernel for the full-on GNU operating system, with no clear target date as to when a full release may be available. I see that Debian have just published a new version of Debian Hurd though, so you're welcome to give it a try.

[Anonymous Coward]

Re: Microkernel

I don't know what the story is with Hurd, but many individuals have written their own microkernel operating systems from scratch. So why it took the FSF so long to get Hurd to work says more about the organization than microkernels in general.

[Henry Wertz 1]

Re: Microkernel

You're right regartding the advantages of a microkernel. But the big disadcvantage is speed. All that jumping back and forth between user and kerne mode and message passing tends to be ruinous for speed.

I'll just note here Linus and Andrrew Tannenbaum had a heavy discussion (that some dubbed flame war but both participants agree was a friendly technical discussion and not personal) over this very issue, back in 1992, there's a wikipedia article about it and the thread was public so I'm sure it's still available. Essentially Linus liked the microkernel design from an aesthetic viewpoint but preferred monolithic from a practical standpoint (and Tannenbaum's Minix is a microkernel, he thought by 1992 making a monolithic kernel makes it obsolete from the get go.)

[Yet Another Anonymous coward]

Re: Microkernel

IIRC Windows NT started as a MicroKernel because Dave Cutler is a proper software engineer

But it had to move graphics into the "microkernel" for performance reasons, then IE, then clippy, then .....

[Someone Else]

Re: Microkernel

Windows NT was no more a microkernel than was VAX/VMS.

Oh, wait...

[Anonymous Coward]

Re: Microkernel

Processors are so fast these days that performance can never be a valid argument against microkernels. Yes, you can improve the performance slightly with a monolithic architecture but the disadvantages are such that we cannot build our computing world around it. As we see almost on a daily basis with vulnerabilities being exploited in all manner of software.

It's time the U.S. government started mandating microkernel operating systems for critical infrastructure. We simply can't go on this way.

[druck]

Re: Microkernel

Any government using Microsoft products has far larger things to worry about than abstract arguments about kernel architectures.

[Anonymous Coward]

Re: Microkernel

I would think that security is foremost on their mind. But maybe I'm mistaken.

[Anonymous Coward]

Re: Microkernel

Now that Linux maintainers are talking about introducing ASI [1], which has a 13% performance hit, we might as well consider switching to a full microkernel.

Most microkernels have a lower performance impact than that (on the order of 5 - 10%).

[1] https://www.phoronix.com/news/Linux-ASI-Lower-Overhead

[Michael Hoffmann]

May I suggest ...

Lennart Poettering?

Sorry, yeah, that wasn't funny even in trolling jest. It's the sort of thing you wake up screaming about.

[m4r35n357]

Re: May I suggest ...

Pardon me, I posted before scrolling to this point ;)

[Fruit and Nutcase]

Re: May I suggest ...

Yes, we may get shafted by the insertion of Agent P by the "blob" - not sure if that would be the equivalent of a Liz Truss

[jake]

Re: May I suggest ...

Except that everybody who might even be close to a leadership roll post Linus is already fully aware that they want nothing to do with the P-thing.

And they especially know they don't want him dabbling in the kernel.

Been there, done that. Never again.

[Anonymous Coward]

Re: May I suggest ...

... It's the sort of thing you wake up screaming about.

Indeed ...

Not funny at all.

Why does no one wonder about "what happens if Leonnard gets hit by a bus"?

.

[stiine]

Re: May I suggest ...

Because its quite bad form to cheer after such events.

[Brave Coward]

Re: what happens if Leonnard gets hit by a bus?

Would be a D-BUS, then.

[Fruit and Nutcase]

Re: May I suggest ...

The bus gets written off

[Doctor Syntax]

Re: May I suggest ...

People wondering about that might lead to buses being hi-jacked.

[GNU Enjoyer]

Re: May I suggest ...

systemd-kerneld

[abend0c4]

Succession

There's a danger in imagining the future of Linux of forgetting that the world will also move on. Up to now, for example, we've had one Linux from desktop to datacentre and a small army of developers producing drivers for everything from domestic TV tuners to Fibre Channel.

I suspect we're going to see an increasing divergence between datacentre hardware and that generally available to the public with I/O in particular more heavily abstracted and quite possibly proprietary to individual bit-barn operators. We've already seen how RedHat is managing source availability for the enterprise market and it's not hard to imagine how that might go if combined with proprietary hardware.

I'm sure there will be a continuing demand for open code that supports a whole gamut of electronic devices, but history is not on the side of there being just one solution to all problems, however convenient it may be. As long as Torvalds is around, the status quo will persist. In his absence, Linux will persist - but perhaps not in the same way and it's perhaps wrong to imagine that it could or should.

[VoiceOfTruth]

Really?

>> It is entirely user-focused and built by volunteers.

So people WORKING at RedHat are volunteers? Come on, now. I know what you're getting at, but this is not true.

[Sloth77]

Rust

" It’s a conservative, risk-averse community, which is a good thing, which makes necessary innovation, like moving to Rust, hard"

Innovation - maybe?

Necessary - definitely not.

[Spazturtle]

Re: Rust

Younger people don't want to learn C when there are better options. If Linux doesn't start moving to Rust then it will run out of maintainers.

[m4r35n357]

Re: Rust

Better tell Linus then. Quickly now!

Those kidz^H^H^H^Hjive talkin' a1 monkeys are sooooo mad for rust ;)

[Spazturtle]

Re: Rust

Linus already knowns, he has talked about this issue multiple times and it is one of the reasons he supports Rust in the kernel.

[Henry Wertz 1]

Re: Rust

Now really. I've used C, I've used Rust, and essentially Rust is like C with all the warnings turned on, 'Werror' (treat warning as error) turned on and 'lint' running at compile time and throwing errors if it finds anything it doesn't like.

Don't flame me too much, I'm pro Rust and I know there's a lot more to it than that.

I'm just saying you coiuld (in theory) run low on C programmers because they prefer Python (I prefer it, although I am aware it's inappropriate for an OS...) but I don't think you'll find a lack of C programmerrs because of Rust.

[G40]

Re: Rust

Tosh. Data to support your claim please

[zootle]

Re: Rust

People said the same thing about COBOL in the 90s. There's still good money to be made maintaining ancient COBOL systems!

[bazza]

Re: Rust

And if you own a COBOL system that you want maintained, you have to pay through the nose to get the time of those very few devs who know COBOL.

It's a warning from history for free software projects that depend on unpaid volunteers. One day, they won't be unpaid, or they won't be.

[jake]

Re: Rust

Yes, COBOL coders can make good money. It's a supply and demand thing. Remember, almost all of the money in the world is handled by COBOL in one way or another. Over the decades, when asked by students, I have always suggested COBOL and/or Fortran as extra languages to learn.

As for the "unpaid" volunteers ... I have been contributing to the FOSS world since before BSD was BSD (and indeed, before Microsoft met an IBM PC). Quite frankly, I have never thought about getting paid for it for one simple reason: It doesn't matter.

Read that again, it's important: It doesn't matter.

I wrote code, created patches, chased down bugs, wrote documentation, and all the other bits & bobs that go into FOSS because I am extremely selfish. I wanted it to work for ME, my way, in my time. Once it worked the way I wanted it to work, it solved a problem that I had, which more than paid for the time and effort that I put into it.

Then I released it to the wild, without caring if anyone else needed it. It's MINE, it scratched my itch ... now, if you have the same itch feel free to make use of my scratching post. No point in you re-inventing the wheel to do the same job ... and better, it frees you up to work on something to fix another itch.

Thankfully, other people have many other itches. In aggregate, over time, and over the generations, we have created something useful. This will continue indefinitely.

[m4r35n357]

That is the EXACT opposite of what is intended by keeping sources together. Do you realize that?

Why don't you try it yourself, and find out how labour-intensive & pointless it is?

How many "specialized" kernels do you think it would take to give equivalent HW coverage?

[Gareth Gouldstone]

Ahem, Unix Wars (1980’s), anyone…?

Linux Kernel wars (2030’s), maybe?

[Gene Cash]

How to tell me you've never compiled a kernel and never done "make config"

It already has exactly that. You can configure your compiled kernel to be "stripped to operate ONLY on a ThinkPad" by selecting only the code you want compiled into the kernel. You do "make config" and hit "Y" or "N" A LOT. For several hours.

[jake]

"For several hours."

Possibly days, or even weeks, depending on how many times you get it wrong doing it manually.

It's a computer, use it as one. First edit the necessary file(s), THEN run make.

_{And even then you'll probably get it wrong ... but at least the logs will point you in the direction of what to fix ...}

[Someone Else]

_{... but at least the logs will point you in the direction of what to fix ...}

Ahh, yes, the joys of deciphering opaque, unintelligible nmake logs written in a 1980's dialect of Gibberish.

Yes, the good ol' days. I remember them well.

_{Or, at least, I think I remember them well. Or maybe just remember them. Anyway...tea time!}

[Rainer]

He has four kids or so?

Can't he train one of his kids?

Surely, one of these would be a capable successor, right?

[Roger Kynaston]

a more heterogeneous environment might be good

I am a linux admin and a refugee from Solaris. At the time at Oracle bought Sun there seemed to be moves to make solaris fully open which, along with BSD seemed like an excellent idea. Of course Larry had other ideas though I know there are project to open source Solris - they havn’t crossed my professional path so I don’t know about them.

Linux, is a remarkable creation but it would be great to see some other FOSS OS out there. Perhaps even a non ‘ix one as well.

Two penn’rth into the void done!

[Doctor Syntax]

Re: a more heterogeneous environment might be good

"Perhaps even a non ‘ix one as well."

Interesting thought. Of course Unix was an excellent model. What would a non=Unix-style OS be like? Would it take the lead from something else? Be totally new?

[Yet Another Anonymous coward]

Re: a more heterogeneous environment might be good

>What would a non=Unix-style OS be like?

It would be like taking VMS and removing the useful clustering features and turning it into NT ?

[zootle]

Re: a more heterogeneous environment might be good

OpenSolaris became Illumos, which is still alive and kicking :)

[CorwinX]

Perhaps they/he should appoint a "Co-Chairman"

Or "Deputy President".

Similar to a Board of Directors.

It's not just about him stepping down - he could suffer a sudden illness or something worse.

There should always be an official 2IC (2nd in Command), be it corporate, community or military.

Primed and ready to step in at a moments notice.

[Richard 12]

Re: Perhaps they/he should appoint a "Co-Chairman"

I presume that they have done.

However it's risky to publicise who takes over, as it makes them a target.

[thames]

It happened before with Python

Let's look another hugely successful open source software projecdt - Python - one of the top languages in use today by any meaningful measure.

It too was started as a hobby by someone in his spare time, in this case Guido van Rossum. This too became hugely successful. Van Rossum also was a "Benevolent Dictator for Life" (BDFL) for his project, and most people could not imagine the project carrying without him.

However, in 2018 he announced that he'd had enough of being in charge of Python and was stepping down as BDFL. He was willing to stay around as a developer working in his spare time on things that interested him, but someone else was going to have to take over managing things because he was tired of being a manager and wanted more time for real life.

And so other people took over and things carried on more or less as before without any real problems. Python continued to go from strength to strength, with continuing development and major innovation and a steadily advancing user base. Van Rossum is still around, but he just works on things which interest him now.

The "irreplaceable man" turned out to be replaceable after all. I suspect the same will be true for Linux. Linux is far bigger than one man, and while Torvalds may be very good as a manager of people (he would have to be in order to be able to bang heads together when he doesn't control their pay cheques), I expect that someone else will be found to step into his shoes when the time comes.

[tamegeek42]

Absolute trash

Every single article by Mr. Goodwins is some kind of click-bait hot take at best. I wish The Register had a feature which allowed me to hide articles by a particular member of the staff.

[Throatwarbler Mangrove]

Re: Absolute trash

Fucking helpless millennials, needing everything to be done for them. Just read the goddamn byline and skip the article, you useless twit.

[Anonymous Coward]

Re: Absolute trash

Look, us millennials are getting old and set in our ways, cut us a bit of slack eh?

[Someone Else]

Re: Absolute trash

OK, Millennial!

[Excused Boots]

Re: Absolute trash

Harsh, but fair!

[Will Godfrey]

Re: Absolute trash

Never mind the kids, what about us baby-boomers?

[Someone Else]

Re: Absolute trash

We still know how to read¹, and have an attention span sufficient to finish the entire article.

_{¹ And... we can even read cursive...}

[trevorde]

Clear successor

Elon Musk - he did such a good job at understanding the Xitter codebase

[jaypyahoo]

BSD exists before Linux and Linux will be here after windows discontinued

[phuzz]

Testing your backups

Whatever gets decided, Linus should take a good long holiday, say six months or so, for doing any kernel work, so we can all see what issues arrive from not having his hand on the tiller.

[Fruit and Nutcase]

Re: Testing your backups

Several months at one of the Antarctic scientific bases in the winter, tasked with counting penguins or something, with Internet access limited to keeping in touch with friends and family

[steelpillow]

Bollocks

> Hope, as they say, is not a strategy.

So curating an expert and motivated community, from which somebody will sooner or later find themselves the lynchpin, is not a strategy, it's blind hope. Fuck me, fuck my socks!

Tip: If you start to sound like some proprietary dot-com's shill, take ten and think about it over your favourite beverage.

[Anonymous Coward]

"Like moving to Rust"

BINGO!

You almost managed a Linux article without it. Good luck for the next one!

[Ian Johnston]

The Linux kernel is already a bloated nightmare, so bug-ridden and insecure that updated versions have to be released - if Linux Mint is anything to go by - roughly twice a month. It tries to do far too much, on far too many systems and is in any case based on a cobbled together minicomputer operating system of 50 years ago.

So yes, well done lads, and thanks, but I find it very hard to imagine it still being a leader until Linus retires, let alone afterwards.

[druck]

If they only did updates once a month, say on a Tuesday, it would only have half the bugs and be twice as secure.

[Captain Zippy]

i mean Patch Tuesday is _right_ _there_.

And it's monthly because people were fed up of MS releasing patches every other day to fix some critical bug in IE.

[druck]

And lets not forget the forced reboot(s) on every one, unlike Linux.

[PhilipN]

Gave me pause

And urgent check on the progress of GNU Hurd as a fallback. 35 years and counting.

[Anonymous Coward]

Power Vacuum

That's the thing that'd lead to the fragmentation of Linux. As soon as commercial concerns see that there's no effective project leadership overseeing the production of useful kernels, they'll have to go it alone and become RedHat-ised, just to survive, or make a switch to something else. If the "usable" distros go commercial, and the free ones are left hamstrung by an under-performing kernel project, that'd be that.

Different projects have different styles of leadership. Some are dictatorial, driven by one person. Some are collegiate. I suspect that switching between leadership styles would be the hardest thing of all to do, and keep the project's coherency together. In human history, there's not many dictatorships that have cleanly transitioned to democracy, or vice versa. Dictators are notoriously bad at retirement planning.

However, this is only software, not politics. I think that - ultimately - because it's LT's project, it's down to him to decide how he wants to retire and what he leaves behind when he does. The rest of us will cope just fine, and be thankful for the ecosystem that's built up around Linux. Should the project go to rack and ruin afterwards or get swamped by commercial interests, there's other OSes to switch to readily available (e.g. FreeBSD).

[Will Godfrey]

So much speculation

Why does everyone seem to think this hasn't already been given considerable thought and preparation.

I don't know if it has or not, but would not be surprised if a transition was made quietly and without any fuss - indeed, for all we know something could already be in progress while the doom-sayers wring their hands.

[jake]

Re: So much speculation

"Why does everyone seem to think this hasn't already been given considerable thought and preparation."

It has. See my first post.

[Uh, Mike]

A. Linus' longevity is a poisson distribution. There's a real possibility that the number of his remaining days has a very small number of digits. Reference: Buses vs pedestrians.

B. Having a stable government doesn't mean you can keep the government stable. Reference: The United States of America.

[Herby]

Sometimes succesion happens.

Take a recent example. The UK had a long serving (70 years) queen. Now we have Charles (the former prince of Wales.

From the looks of it we need a "penguin prince" that has been anointed from the Emperor Penguin himself. Sometimes this works, but sometimes the heir apparent goes away before the transition (e.g. Berkshire Hathaway).

We may never know the succession to the Penguin, but it might be a good idea to think about it, and given them (whoever they may be) the proper training to take over the crown in the style of Linus. Hopefully the successor will have the same (proper) temperament as the original.

Just remember when we lost DMR!

[AndyJWard]

I think the most likely way forward when Linus steps down is a small comittee where he has a say in its members. Then it's effectively "hands off" and let it run. It woun't be entirely smooth, and there will be squabbles, but for the devs that respect Linus and Linux for the magnificent creation that it is, I have faith that it will work out.

[Fruit and Nutcase]

AI+Blockchain

I'm surprised someone hasn't scraped the Linux kernel mailing list and come up with a Linus Bot

[boatsman]

all things that could go wrong when Linus is gone, can also go wrong right now.

and thats all there is to it.

splitting, flamewars, fracturing ... etc etc etc.

[Fruit and Nutcase]

Re: all things that could go wrong when Linus is gone, can also go wrong right now.

Like this

https://www.theregister.com/2025/08/15/sad_end_of_bcachefs/

[Timop]

It is not the things one is able to spot but things that might slip through that matters more than one wants to admit.

And this is why I think this has a point.

The profit motive is out there. Any company would extract everything possible for next quarter profits and left smoking ruins behind. Unless there are people smart enough to spot it and brave enough to call it out.

Imagine for example someone with conflicts of interest. Get sacked or do the free work part just as we say. All they need to do is align incentives and rest happens probably automatically.