Windows to become emulation layer atop Linux kernel, predicts Eric Raymond Open-source software advocate Eric S Raymond has penned an argument that the triumph of Linux on the desktop is imminent because Microsoft will soon tire of Windows. Raymond's argument, posted to his blog late last week, kicked off with some frank admiration for Windows Subsystem For Linux, the tech that lets Linux binaries …
It's not a trial of anything. Earlier, Microsoft asked Linux programmers if they were interested in Edge. Only after getting a large positive response did they then announce that they would port it to Linux. The reality is that according to Stack Overflow surveys, development on Windows has fallen to below 50% for a few years now. Hence, not having Edge on Linux would lose Microsoft more mindshare with developers (a reason they bought Github). Edge on Linux helps ensure that web apps are tested to run correctly in Edge. That's it. Raymond's assertion that the only reasonable explanation is that they intend to run all of Windows on top of Linux is absurd. Heck, the success and continued improvements to WSL show that if they intend anything it's to run all of Linux within Windows.
I drink Coke, never Pepsi because Pepsi is an imitation who wants to be as big as Coke. I use Chrome and not Edge because, Edge is an imitation who wants to be as big as Chrome. The whole Windows should be rewritten using the more efficient Linux then, I'll consider it.
From a cashflow point of view, Coke has been the smaller player for some time. Although to be fair that's because Pepsi is pretty much the world leader in potato crisps... (Walkers? Pepsi. Lays? Pepsi. About the only brand they don't own is Golden Wonder... so far)
Yes, it probably doesn't require hardly any work at all. I wouldn't be surprised if they didn't have an internal Linux build the whole time.
Microsoft have finally woken up to the truth of their situation. Windows is just a poor, proprietary client for a Unix / Linux world, and has been for years.
for the Linux Kernel team should this happen.
MS will want more and more control of the underlying layers. I fully expect that the core of Kernel Devs will shrink and be replaced by MS people. Linus will be pushed out and the job will be done... well part of it.
I see their next task as taking key parts of the kernel out of GPL. Us refuseniks will still be able to build a working kernel but more and more key elements will be for MS licencees only. As they extend this, there will be parts of the open source kernel that is simply left behind to fester and die. A group will be formed to keep it alove but their job will become harder and harder and eventually they'll give up from lack of funding.
Then Redmond's job will be done and we'll all be beholden to Microsoft (as well as paying them money for patches). The only place left to go will be BSD.
Yes, this is a worst case scenario but...
It is up to everyone to stop them taking over and bleeding it try.
"I don't see that happening any time soon.
Never will. This would open Microsoft up to a code auditing lawsuit specifically for copyright infringement.
SCO vs. IBM proved that when it's time to dig down, they'll step away and step away fast as they left MS_PROXY SCO out in the wind, MS didn't make a peep. It comes down to answering who pay$. If Microsoft finds an infringement, who pays? But if Microsoft is found infringing, that answer is blistering obvious.
You could argue they're poisoning the well, but they too now go to that well, so...
I think he means that they will replace GPL modules with proprietary ones, and then stop supporting the GPL ones (or even breaking things so the GPL modules are no longer compatible). Not quite the same as "taking components out of the GPL", but has much the same effect.
In a way, although I don't think it's the most likely option, it wouldn't surprise me, as it would just be the old "embrace, extend, extinguish" playbook all over again.
That said, I don't think they would actually replace Windows with Linux unless Windows started to lose money of hand over fist.
I don't think they would actually replace Windows with Linux unless Windows started to lose money of hand over fist.
They haven't made money from Windows for years. Their core businesses are "Office" and "Office 365" - software as a service was seen as the approach that would rake in zillions, but "365" is anything but successful.
GPL or not is irrelevant. The issue is manpower, and he who can bring the biggest amount of manpower to bear will end up controlling it, even if it is open source. Linux has got this far merely because Linus has been at the head of a small army of the willing who have been happy to have him in charge. Someone comes along with a bigger army of devs, they can take it in a direction that many users may find compelling but is not easily replicated by the original team.
Look at file systems; Linux hasn’t got anything better than ext4, whereas ZFS is available to anyone who can stand its license, Apple now has afs and even MS are waking up to replacing btrfs.
Look at how RedHat have got systemd into more or less every distro by having control of Gnome and making that require systemd, which they also control.
I would agree with you, but they cannot take components out from under the GPL, not legally at least unless the person who wrote the code and all the people who modified it agree. I don't see that happening any time soon.
I'm sure Microsoft are eyeing what Red Hat did with systemd and wondering how to apply the same strategy, except with non-GPL software. Google did the same with Android: a free, libre, kernel, but Google Play Services binds people nicely into the Google ecosystem.
Windows as a proprietary 'Google Play Services' layer on top of the Linux kernel is entirely possible, and reduces the amount of money needed to develop and maintain the operating system. Linux will have 'won', but if the end-user experience requires the proprietary layer on top, it isn't free.
Google/Alphabet appear to be keeping their options open so they have the technical ability to replace Linux with Fucsia (or similar), should the need arise.
"It's getting hard to find a distro that doesn't use systemd."
Slackware still works quite nicely. Perhaps your definition of "hard" doesn't match mine?
"In consequence it will be increasingly common to find applications assuming that one of other of its tentacles is available for use."
I have heard many people make this claim. However, io date I have not found that to be an issue even one time. As I said, Slackware still works quite nicely.
GhostBSD has a telegram group. You can download an .ISO file and burn to USB Flash drive and test boot up the live image on your computer and see if you like this BSD. Take a few minutes and check it out as a replacement for your daily driver.
No SystemD, use OpenRC to start systems. Based on FreeBSD 12.1 Stable but from GhostBSD controlled repository. Gnome HOSTed BSD, uses MATE desktop.
Thanks for the consideration of useful BSD, if you don't like SystemD.
Yes, I use Manjaro Linux installed to SSD, BTW. ;>) and Puppylinux frugal install from a USB flash drive. So not Linux adverse here. Use the OS that works for your work flow needs.
Yes, Jake. I know about Slackware. Personally I use Devuan. But these are not the mainstream distros. It's more likely that anyone developing mainstream applications will be using Ubuntu or one of its derivatives - it seems to be getting quite common to see .debs offered for Ubuntu but not Debian.
Oh. I didn't realize that a distro had to be mainstream to be counted. My bad.
Remind me again what "mainstream" means? I'd hate to make that mistake again.
As a side note, converting between the major distribution formats is fairly trivial ... there is nothing magical about a .deb or RPM ... or a tarball, for that matter.
Clearly, Sir, you are absolutely correct. Despite being in widespread use since at least the 1970s, the phrase "my bad" has no place in the English Language. Here's a list of 500 of the posts made to ElReg containing the phrase during the last decade plus, obviously they all need downvoting posthaste! I assume you will get right on it.
During the meanwhile, we must figure out exactly where the English Language must be frozen in time in order to appease you ... I assume that would have to be prior to the works of one William Shakespeare, who wrote in Sonnet 112:
Your love and pity doth the impression fill
Which vulgar scandal stamp'd upon my brow;
For what care I who calls me well or ill,
So you o'er-green my bad, my good allow?
Ah, well. All the world's a stage, and everyone a critic ... WAIT! The word "critic" was a new term in 1590s English, so I guess that's right out, too ... Mea Culpa. (Presumably I can use the Latin version of "my bad" without fear of your wrath?)
So if we use the Bard as our cutoff point, we're essentially left with Everyman to work with, unless we want to go back to Chaucer. Alternatively, we could look to Yola and Fingalian. A third possibility is is going the Aenglisc/Anglish route and eliminating all words and syntax of Latin origin. This last might be the easiest as many have already made inroads.
The beauty of Linux (and open source in general) is that people can fork things if they do not like the direction of a project.
Its not like Windows or Mac where you could be actively prevented running/changing a component of the OS.
There are several systemd free distros..
- Devuan (debian based)
- Void
- Slackware
- Gentoo
- GUIX
- Artix Linux (arch based)
Probably many others.
Personally for all its faults I have got used to it now and can see advantages compared to past init systems (I quite liked upstart )
I have got used to it now and can see advantages
Same here. I particularly like the way that when I accidentally type 'sudo systemctl disable someservice' and then also accidentally type my password, it saves my from myself by giving output that seems like it's disabled someservice, but not actually disabling it. This intuitive feature has saved me from having free memory on a bunch of occasions.
That's unfortunate. If the command does not remove the symbolic link for starting the service at boot as indicated in the documentation that's a bad thing. Perhaps a bug have been raised?
Personally I like it as well. I spent a few hours reading up on how it works and investigating the commands and find it very practical. The unit files are great as well.
Yeah pretty much this. I gave up filing bugs against pottering garbage sometime around ~2015 maybe.
The one that sticks in my mind is the time I was told that the reason why pulseaudio uses a bunch of processor time redundantly doing software mixing even though I have dedicated hardware specifically designed to do that exact job is because piss off you dumb user you shouldn't have pulseaudio installed if you have a multichannel sound card, duh. This despite the fact that it's installed and active by default with most distros, and that some software only targets pulseaudio *cough*firefox*cough*.
Somehow that bug wasn't the straw that broke the camel's back, but I don't remember the details of the one that did.
But I'd be happy to file a new bug for pottering to ignore if that's what you'd like me to do. My consulting rate is $100 per hour or part thereof. I estimate that I'll be able to collate a bug report for you in 10 hours or less. Please send through your billing address and a note saying "I wish to engage your consulting services and accept the rates you have advised" and I'll get right on it.
Ha! Mine was "network sinks randomly use 100% CPU when idle" to which the response was "why are they idle? Audio subsystems are for playing audio". Like "how dare you install pulseaudio then go to sleep?"
Since someone else took over Pulseaudio it has actually become quite good.
Likewise systemd isn't terrible to those of us not emotionally wedded to a bunch of unintelligible shell scripts. So long as we don't attempt to report bugs,
How dare you fall asleep while listening to an audiobook! :O
emotionally wedded to a bunch of unintelligible shell scripts
Ha. I just wrote a rant about that minutes ago. It's not emotional at all, it's a simple case of not seeing any need for a new, unnecessary, less-flexible domain specific language when a generic one that everyone already knows will do just fine.
They're only unintelligible if you don't know bash. And if you don't know bash then you're not qualified to be administering a Linux system.
You're right though, the unit files are great. I've earned thousands of dollars just reading the documentation and trying to figure out which particular magical keywords need to go in which particular magical files. And I'll earn thousands more doing it all over again, because I can't remember any of it due to the specificity of the DSL and the infrequent schedule on which I'm forced to use it. It really is great.
But I think the best thing about unit files is that they save you from having to make the effort to become competent at a well-established, ubiquitous, super simple, massively documented shell scripting language that anyone administering a Linux system should already know and use every day and which has 30+ years of testing and use and bugfixes under its belt. That sounds hard.
This post has been deleted by its author
"seems like it's disabled someservice, but not actually disabling it. "
"At the very least systemctl should throw an error giving some clue as to WHY it is not disabling a service."
I believe the initial complaint is about what it does do when the service exists, namely it disables it. Disabling it means that it removes the service's links which determine whether that service will be started under conditions like a reboot or a change in level. It doesn't stop a service. The reason I think that's the complaint is the mention of high memory usage; I'm guessing the problem here is that the disabled service is still running because systemd hasn't had cause to restart services. The command to stop a service is "systemctl stop someservice" although you can combine them by specifying disable with --now. I have retrieved this information from systemctl's man page which includes a warning that disable doesn't mean stop.
Now it may be argued that a command called disable should both disable the service from being automatically started and stop it now, but the documentation is clear that it doesn't do that. I am not here to praise systemd; I don't much care whether you like it or hate it. There are real bugs in it which are annoying and I'd like to see patched. Still, if the problem you have is what commands look like, you can't really call that a bug. Disable disables and stop stops. Want to create a disablestop or modify disable? Fine with me, but file it under changes and not bug fixes.
I believe the initial complaint is about what it does do when the service exists,
Yep
namely it disables it
Nope.
Disabling ...(snip)...doesn't stop a service
Indeed. And as a bonus feature it also doesn't disable if it doesn't feel like it!
The whole "we decided that 'disable' doesn't mean what the word 'disable' means" thing is also a problem. But it's not the problem I'm talking about - I can read a (needlessly huge because everything has been needlessly combined into one command which is needlessly named extremely similarly to the preexisting 'sysctl' command) man page.
The problem I'm talking about is when I disable containerd, and then later I reboot, and then later I notice containerd is running, so I disable containerd even though I'm pretty sure I already did that a few days ago, and then I kill containerd, and then disable containerd again just to be sure, and then reboot, and then notice containerd is running, and then consult man to make sure that I'm using the right command, and then disable containerd, and then kill containerd, and repeat that sequence 47 times, and finally get tired and just uninstall docker and containerd because I'm not really using them on that machine anyway. And then when the same thing happens for mysql, rather than going through the loop of frustration again I just install devuan and the totally-not-a-bug magically goes away.
But I'm sure that somehow that's a bug in containerd and mysql, or maybe in the kernel or microsoft office. I'm sure it's not a bug in systemd. Or that it's my fault for expecting ''disable' to actually do the thing it says it's doing in the output and that man says it should do. It's probably that I'm a dumbass and not that this new piece of bleeding edge software written by someone with a track record of writing dogshit, is dogshit.
I'd look into it further, but I'd rather pull my teeth out using something sharp and rusty, and the systemd guys still haven't sent through that billing info.
The command to stop a service is "systemctl stop someservice" although you can combine them by specifying disable with --now
As an aside and while I'm feeling ranty, I love that "disable" doesn't mean stop, but "disable now" does. Yay for consistency! This is not at all completely backwards and self-contradictory. It's simple and intuitive: the default behaviour is to defer disabling until reboot (because otherwise you'd be doing it now). But there's a helpful option to tell it to disable now, which stops. But disable doesn't mean stop - there's a special note in the man page telling you this, duh, you dumb user.
"Of the two which is going to be used fro application development?"
Quite honestly. it shouldn't matter. make install should work on any distribution put together by sane maintainers. If it doesn't, I humbly suggest either the code you are trying to install, or the distribution you are trying to install it on has been gotten at by Marketing with the explicit intent to tie you to whatever their flavo(u)r of the month is.
"It's getting hard to find a distro that doesn't use systemd. "
Then you can't be looking very hard. As jake says slackware is free of the systemd taint and so is PCLinusOS. I don't know if Patrick Volkerding has any thoughts on moving to systemd but TexStar will definitely not.
It is true that a lot of distros acted like lemmings and adopted systemd but this being Linux there is always a choice.
"It is true that a lot of distros acted like lemmings and adopted systemd but this being Linux there is always a choice"
I don't think that's being entirely fair not least because (and much to my surprise) it was Debian that voted to adopt systemd and that decision then affected all the downstream distributions. I wish it had been otherwise.
I will admit that I am not a Linux expert, so I'm asking in all seriousness: To all the posters saying there are plenty of distros that don't use systemd -- how many of them are ones that you'd be happy to give to your non-techy parents or your great-aunt Lucy? Because, as I've said before, Linux won't be mainstream until those people can just sit down and do the things they want to do. Ignoring for the moment the stories we all know of people who are completely atechnical ("Point and click? What's that?"), most people who learn to use Windows once can - for the most part - go anywhere else, sit down at a computer, and do their thing. With a plethora of distros, desktops, etc., Linux doesn't have that, if you will, enforced simplicity.
Now, if -- and I stress *IF* MS were to be able to use its well-known "embrace, extend, and extinguish" strategy to, as has been suggested, create a familiar shell that, e.g., uses systemd, the fact that it acts like the Windows that average users (Hi, Aunt Lucy!) are used to implies that it will become the default distro for new computers sold to private consumers, businesses, libraries, etc. This puts other distros where they have always been; as also-rans used primarily by the techy/enthusiast markets.
And, as we have seen multiple times in living memory, the user interface - the look and feel -- are cooyrightable, limiting how closely the Winux interface could be imitated.
Which puts us back where we started, only this time with a Windows-branded Linux as the proverbial 800-pound gorilla for the consumer market and numerous (relatively) little-used forks sweeping up the leftovers.
What am I missing?
How about the fact that Apple lost the copyright case?
"To all the posters saying there are plenty of distros that don't use systemd -- how many of them are ones that you'd be happy to give to your non-techy parents or your great-aunt Lucy?"
Ignoring the implied ageism - I may well be older than your parents, I did provide a couple of cousins with Zorin but that was a long time ago.
But let's qualify this: statement:
"most people who learn to use Windows once can - for the most part - go anywhere else, sit down at a computer, and do their thing"
Which Windows? The Windows UI has changed a good deal over the years. How many complaints have there been here over just that thing. And not only the OS, the applications as well - remember all the complaints over the introduction of the ribbon? In fact if you sat down in front of my laptop you might wonder for a moment which variation of Windows it was running. W95 and KDE both seem to have started by adopting a good deal of the Unix CDE interface and IBM's CUI so there are quite strong resemblances between them. and other Unix/Linux window managers. In consequence the differences between any given Windows version and many* Linux UIs is not really greater than that between different versions of Windows.
* There is a good deal of flexibility but generally KDE based implementations usually default to having the main menu pop up from the bottom left with the panel (task bar) etc on the bottom. I'm not sure what the most common Gnome layout is these days but most distros using it used to put that sort of stuff at the top of the screen. But hey, if you want your task bar down the right hand side that's possible.
>how many of them are ones that you'd be happy to give to your non-techy parents or your great-aunt Lucy?
Assuming Aunt Lucy is already using a phone or tablet then they're already using a flavor of Unix, be it BSD based for Apple or Linux based for everything else. The rise of 'apps' culture has also shifted the platform that people work on to the web browser so even if she is working on an older Windows system she's likely to just bring it up, open Firefox and that's it for just about everything.
Where Microsoft is going is to wall off the business space with their own flavors of groupware that are tightly integrated with their Office products (which are now also web based). The strategy is obviously to capture and retain the business market -- its huge, it has deep pockets and it is resistant to change ("there is no alternative"). The OS these applications run on is irrelevant and if Microsoft could gradually phase out Windows as we currently know it then it would save them huge amounts of development and support costs.
The times I've tried a BSD felt like death by 1000 paper cuts. Not because it was different to the GPL equivalents, of course I expected that. Lots of details didn't quite work as advertised.
Don't ask for details though, it's been a few years.
I think you'll find that there is a lot of 4.3BSD in SunOS5.0 (AKA Solaris), by way of SysVR4.
AIX had boatloads of 4.2BSD in it, with a hefty pinch of early SysV.
The Mach kernel itself started life as 4.2BSD, and userland in MacOS and NeXT is clearly BSD.
"I think you'll find that there is a lot of 4.3BSD in SunOS5.0 (AKA Solaris), by way of SysVR4."
I'm probably wrong, but I was under the impression that Berkley itself was heavily funded by just about all commercial UNIX operators, thus making BSD kind of everyone's all at once, not just Berkley's.
I wish to point out that AT&T sued U.C. Berkeley because 6 files were co-opted from AT&T Unix into BSD to make it into a full fledged operating system. However, it was discovered that AT&T was removing the copyright notices from the BSD source files and the man pages in violation of the license. So they settled out of court and the agreement was sealed. This was back in...the early 1980's I think. Fast forward to SCO vs. IBM. That sealed agreement was unsealed and posted to the internet for all to see. As far as I can tell, there has been no final ruling on the case. The last update was in 2019, which means that the case is still ongoing...17 years later.
"I think you'll find that there is a lot of 4.3BSD in SunOS5.0 (AKA Solaris), by way of SysVR4."
It's a long time back to stretch my memory but surely System V, like System III was AT&T. I suppose there was a System IV in between but I never used that.
AFAICR the original AIX was by Interactive who also did the V7 port to Z8000, Onix, wich was the first Unix I used - and that showed in the AIX I encountered later.
There was a lot of code-swapping between Berkeley and AT&T until AT&T's lawyers noticed that UNIX might be a money maker.
The IBM/Interactive Systems Corporation kludge included IX/370 and PC/IX (which were odd-ball variations on the theme, to say the least) and AIX Version 1 (which was BSD and SysVR1 & 2 based and ran only on the RT PC) ... All later versions of AIX were pure IBM. And showed it.
There was no System IV ... there was no System I or System II, either.
Hold on a sec please.
I grew up with xenix on a tandy 6000. I came into this on the cheap bastardized side. Then I moved into SCO Xenix, Unix on a variety of architectures. Hell, I played with the first NCR tower. Then I ended up on some decent sun gear. Then I put a lot of Linux boxes in.
My boss came to me and said there was a printer on the production side, a multi-million dollar printer, that would freeze up due to a random packet.The manufacturer said as long as you can't guarantee that particular packet doesn't show up from the network we will not support that printer and you will lose millions of dollars.
My boss, a brilliant tech / marketing guy, spent a day or two researching open BSD networking rules and gave me an example of how I might save the printer.
I spent a couple of more days thrashing out the plan, the various networking cards, the various networking rules, and came up with the solution. I set up two failover boxes that filtered that network and isolated the production from everything else. I tested for 10 times the possible throughput. And we dropped it in.
3 days later we found that God damn packet. We figured out the random equipment that triggered the horrible reaction. This became the default plan of how to isolate the network on the production side. It also became the logging system that was able to track every brownout that ever happened and focused on every malfunctioning piece of production. I could roll back a week worth of data to analyze in Wireshark and say that piece of equipment failed at that point. It was amazing, and it was one single box.
So let me please sing the praises of open BSD and it's ability to filter and save a network production system. It took a week and it was amazing and it probably lives on today.
Now back to Eric. I love the guy. I went to a lecture that he gave around 1985 ish. It was about porting applications between various unixes. He was amazing. He was informative and educational and just had the right amount of aww shucks as he shuffled around.
Let me tell you about the shuffle around. He's got problems with his legs and feet and his club footage. I have problems with my back, hip, legs and I shuffle around the way he did as he was suffering out on the lecture stage. That guy had to deal with some serious physical disabilities and he was not letting it shut him down. He was also deep into martial arts training at that time. He's brilliant in many ways.
At the end of the lecture I told him I was working under SCO Unix and I did not have the korn shell that he was talking about. He arranged the tape for me. It was up to me to port it with the various compiler uh-ohs. It was awesome.
Any OS responsible for horrors like systemd deserves to die.
That was Poettering, not Linux. And rather than 'die' I was thinking of something involving a wrought iron fence and a rubber chicken, but oh well...
/me typically uses Devuan whenever I need Linux
"You didn't mention a video camera. I'll be angry if this happens and there's no video camera involved."
You'll probably want a second video camera too. You know, to like, er, video it as well. The first video camera should be one of those early VHS "portable" ones, for extra effect during the insertion.
GhostBSD has a telegram group. You can download an .ISO file and burn to USB Flash drive and test boot up the live image on your computer and see if you like this BSD. Take a few minutes and check it out as a replacement for your daily driver. No SystemD, use OpenRC to start systems. Based on FreeBSD 12.1 Stable but from GhostBSD controlled repository.
Thanks for the consideration of useful BSD, if you don't like SystemD.
Yes, I use Manjaro Linux installed to SSD, BTW. ;>) and Puppylinux frugal install from a USB flash drive. So not Linux adverse.
Linus will be pushed out and the job will be done... well part of it.
If he is forced out it will be fascinating to see Linus forking the Linux kernel to keep a free version.
But I doubt that this "MS take over" will happen - there are too many others who depend on the Linux kernel.
The creation of a common global infrastructure with no dominant controlling* entity is more likely to result in a stand-off where they all continue to push their code into linux as now and benefit from the volunteer effort as well. Any attempt to take control (by any single big player) would create a temporary alliance of convenience to stop them.
Of course, this doesn't preclude a cartel from forming.
*Linus has remained in the driving seat only by consent of the developers and not getting into commerical competition with the big players.
The only place left to go will be BSD.
Apple has been using a fork of FreeBSD's userland for years. Hardly "taken over".
I think you fail to see the clear advantages here, with a choice [or not choice] to run some kind of Windows emulation on top of Linux (I'd like mine working with FreeBSD as well, please, and no required systemd hooks, k-thanx). But such a move might also explain why 10's of thousands of lines of code were recently attempted to be dropped on kernel devs for updates to kernel-side NTFS support... [not sure what exactly happened with that, though]
I'd actually PAY MONEY for something _similar_ to Wine or Steam's game emulator, running on top of Linux.
I just hope Micros~1 won't FORCE A WIN-10-NIC APPEARANCE (or 'Settings' behavior) on the userland side... yeah I wanna keep my MATE desktop with 3D SKEUOMORPHIC LOOK. And no ads, either. And NO SPYWARE. [yeah good luck THAT happening, I bet]
Icon, 'cause, FreeBSD
(there's no systemd in BSD - YES!)
I wouldn't be so sure that Microsoft wants to "take over" the Linux kernel. This might just be the commoditization of kernel functionality. Does it make sense for Microsoft to throw money at something for which a superior solution exists that's free? It could pour a fraction of what it spends on the Windows kernel, file systems, etc. into Linux an and even LSF and probably still be considered a generous sponsor. Redmond could then spend its money building up its Azure stack, developer tools, and enterprise stack (Exchange, SQL Sever, etc.) which are more profitable and have more growth potential.
I don't find this any more evil or suspicious than what Amazon does with taking things like Redis, MongoDB and Elasticache, forking them internally to add scalable storage and some management sweeteners, and charge a lot of money for it. If Microsoft really does continue its trend toward contributing to the kernel, supporting its tools on Linux, etc. one could make the argument that they are better open source citizens than Amazon.
I don’t see Microsoft doing this. There’s far more to it than simply having Win32 on Linux. MS aren’t about to walk away from their decades of Windows (kernel, services, etc) and try and do it all over again on Linux.
For example, we’re highly unlikely to see the vast mass of code that depends on COM being ported over to D-Bus...
MS aren’t about to walk away from their decades of Windows (kernel, services, etc) and try and do it all over again on Linux.
They will. It'll be just like the transition from OS9 to OSX - different underneath but with similar shiny on top. The user experience will be mostly the same, except that their machines will be more stable, more secure, and a lot quicker (the NT kernel is a horrible, slow, tangled mess, similar to the obscenely convoluted OS9).
Apple jumped to BSD when they could no longer support the convoluted mess of kludges that made up OS9. It really hurt them, and MS will feel similar pain when they bite the bullet and finally accept the inevitable. There will be many job losses at Redmond.
M$ have finally woken up to the fact that their OS is just a poor client for a Unix-based world. Just like Apple, they'll keep their app layer looking like Windows, but everything beneath will actually work properly for the first time.
I was concerned when MS were allowed to contribute to the Linux kernel, but when the code was scrupulously analysed and found to be clean, most of us welcomed the donations.
It's a strange world, and it's getting stranger all the time!
I suspect that Windows might end up being built on top of an increasing level of Linux code Wer have honestly. Windows PE may be where it starts. The thing is Windows has not made Microsoft money directly for years, being revenue neutral on the whole and more of a promotional tool to gather data (the biggest industry in the world) and force people to use their cloud offerings (surely the next anti-trust case - although there is little chance of any sympathy for any of the big players - at some point or another they are all a bit shady!) I am guessing that the use of open source to provide free development will help reduce costs, increase compatability for Linux et al All good really. I personally welcome this move. It is not too disimilar to what Apple did with BSD after all.
I don't see it happening. Windows might follow MacOS and become a BSD distro... of course, back at the time of the Windows 2000 Professional source code leak, it was noted that Windows already has a few pieces of BSD in it, but clearly if Windows were to be a layer on top of something, making that something derived from BSD would give Microsoft more control.
replacing the NT Kernel with the Linux Kernel would cost a lot in development and make no extra revenue.
not necessarily. All you would really need is a single translation layer library with all of the userland stuff running on top of it. As long as the USER32.DLL code (and related) can use glx extensions, AND the various video drivers support it properly on the back end, shouldn't be too hard.
Basically, what Wine attempts to do. Sadly, they're MISSING so many important features and compatibility that I can't get anything I want to run under Wine... but if Micros~1 does it, even just FIXING THE WINE STUFF [without b0rking the desktop settings of the user that runs it, AND allowing both 64-bit and 32-bit applications to run at the same time], shouldn't take too long.
Seriously, it would make good economic sense, NOT having to maintain the kernel and drivers.
And, with open source kernel drivers more or less REQUIRED by the GPL (i.e. re-compile driver for updated kernel config), then guess what? NO! MORE! STUPID! KERNEL! DRIVER! SIGNING!!!
Or at least, it *BETTER* not happen to Linux!!!! (it would take "tainted driver" and "tainted kernel" and re-define it in the most HORRIBLE way possible). I _REALLY_ _HATE_ that "tollbooth" that requires you pay Micros~1 to sign a driver, even if you want to GIVE IT AWAY FOR FREE. and it's such a FALSE sense of security, too.
> > replacing the NT Kernel with the Linux Kernel would cost a lot in development and make no extra revenue.
> not necessarily. All you would really need is a single translation layer library with all of the userland stuff running on top of it.
They've gone to a lot of trouble to build WSL, which translates Linux syscalls into Windows syscalls.
Why would they do that, if the plan was to do it in the opposite direction?
You’re out of date. WSL is now a hyper visor hosting a full Linux kernel with some nice integration features.
MS are merely the latest in a long line of outfits to implement the Linux system interface; first there’s Linux itself, Solaris, FreeBSD, BlackBerry10. It always runs into too many problems of nearly-good-but-not-good-enough.
A cutoff point where using Linux costs less than NT kernel development...
yes, I've been thinking that maybe that cutoff point is NOW...
Also worthy of mention though: X11, and *NOT* WAYLAND!!! If Micros~1 requires Wayland for "Windows on Linux" it would be YET ANOTHER example of "getting it wrong".
And, also important: NOT take over the desktop to run Windows applications. Let them run SEAMLESSLY on your Mate desktop, for example.
and wouldn't it be nice if "the registry" were stored in "~/.config/Windows" as the Open Desktop might suggest, either in XML or plain text "INI style" form, using actual files and directories? Especially NOT using a "Big ol' BLOB" that you can't easily back up... (whereas tarballing a '~/.config/Windows' directory would be trivial)
(I'm surprised nobody brought these up, yet, or did I miss them?)
replacing the NT Kernel with the Linux Kernel would cost a lot in development and make no extra revenue.
Having just done some of that recently myself, it's surprisingly simple. The development costs will be relatively trivial, and Win 10 will just become an application layer on top of a Linux OS. It will be largely browser-based (because most of the code already exists - see "Edge" on Linux!).
Am I the only one watching all this Linux/Windows merging from Microsoft and thinking "Oh please no"?
I switched to GNU/Linux over 10 years ago (In three years... twenty years go :-O.. wow I'm getting old). Unfortunately this means my mind is still very much "I want GNU/Linux to just be GNU/Linux".
I really don't feel I want Microsoft code in my operating system (Even though, Microsoft have successfully shoved a ton of unchecked code into the kernel, .NET (mono) onto most GNU/Linux installations).
In my view, I don't see Microsoft essentially taking control of Linux as a "win" for GNU/Linux. If anything it's a loss. The thing we strived to be different from, we end up becoming them.
I'm sorry, but I don't want GNU/Linux to become Windows, or visa-versa. I want GNU/Linux to be GNU/Linux.
How long until something like HDCP is implemented in the kernel, and DRM becomes a standard, built-in part of GNU/Linux? I don't want the DRM and other licensing nightmares that will follow with the Windows merging.
At that point, GNU/Linux stops being a "Free as in Freedom" OS, and it is no longer truly "GNU/Linux".
Free as in Freedom please, not Free as in Windows.
/end rant
Currently, schmerently. If it is going into the mainstream kernel, it will be given the fine-tooth by non-Microsoft employees until roughly the heat death of the Universe. There will always be a subset of the FOSS set who don't trust commercial code at all, and will make every effort to expose any transgressions (real or imagined) attempted by commercial outfits.
Yes, it will be nice to get a look at some of the stuff Redmond is capable of.
From what I have heard some of the Windows stuff may be a little shoddy but I bet that anything exposed to the rigours of an examination of any MS contributions to the Linux kernel would have been gone over with a fine tooth comb.
We shall see.
"Microsoft have successfully shoved a ton of unchecked code into the kernel"
No. If anything, the Redmond contributions to the kernel have been checked more thoroughly than contributions from any other commercial entity. Not only has it been vetted by Linus & the rest of the Kernel dev folks ... it is also very carefully eyeballed by a bunch of fanboi hangers-on, each eager to make a name for themselves finding bugs/holes/backdoors/other exploits in MS submitted code.
I didn't downvote you. But I disagree. The big gain here I think would be that printers, scanners, and a multitude of other useful peripherals might come with drivers that actually work in Unix. It's all very well to babble about SANE, WINE, native unix drivers. My experience has been that they rarely work very well (if at all) without many hours of effort on my part. And often not then. In my experience, the only things beyond mice, keyboards, and monitors that one can count on working out of the box with linux are storage devices or devices that emulate storage devices.
Also, if I understand ESR's argument -- which I may not -- Microsoft simply won't see much profit in screwing with the Linux kernel. Money in, no money back. Maybe they simply won't bother.
Remarkable. Have multiple boxen here with very new peripherals connected to them... Bought new colour laser and graphic tablet just yesterday for working with DICOM. And it just all works. Could it be that the biggest issue in IT is located between keyboard and chair?
We had to deal with some Wacom graphics tablets at work recently. Yes, they work, just (that's "work, just" and not really "just work"), straight out of the box on Linux (which is still pretty good), but good luck with trying to get any of the custom settings working (eg, stylus responsiveness, etc).
They offer a supposedly slightly more advanced driver for download, but it's some nasty tarball that you have to run (as root) each time you want to try to use it, and, as far as I could tell, it then completely nukes any chances you had of going back to the default Linux drivers afterwards.
I like Linux, especially as a server OS, but there is a good reason why some people spend money on shiny Apple gear if their intended desktop uses are slightly removed from the purely vanilla.
(And now that Apple has absorbed CUPS, that at least has the useful side benefit of improving Linux support for printers, too. Maybe Apple could be encouraged to work more closely with SANE as well?)
IMO that hasn't been an issue since the 90s when "WinPeripherals" were the rage among rubbish mom & pop computer stores, and brands like Packard Bell. Your 3Com NIC, USR modem, and PostScript printer worked in Linux without a hitch; cheapskates got the performance and compatibility nightmares they deserved.
I see it differently.
Microsoft could be making more money from Office than from Windows. From their POV, it could be better not to have to worry about the underlying OS and concentrate on selling the more profitable Office and server apps.
From that perspective, having "Linux inside" makes a lot of sense. The other advantage is that Linux users become a market for Office apps, which nicely handles all those businesses (well, only a few, really) who have transitioned from Windows to Linux on the desktop.
>Microsoft could be making more money from Office than from Windows
And they make more money from the cloud than anything.
Imagine if there was no windows desktop and the only way to run your windows apps was on Microsoft cloud then not only do Microsoft get to tie all their users into the cloud, all the publishers of Windows(tm) software become app-store victims just like on iPhone
Even though, Microsoft have successfully shoved a ton of unchecked code into the kernel, .NET (mono) onto most GNU/Linux installations
I call FUD on the first part, and "only if you let the installer do it" on the 2nd. Last I checked, on Devuan with Mate, there's no mono or ".Not" core.
I was VERY angry about 10 years ago when I discovered that gnome installer on Debian was including "tomboy" (a worthless postit note application that I'd never use) in its top level gnome-desktop (I think that's right) package, which dragged in MANY MEGABYTES of MONO CRAP to satisfy the dependencies, JUST for tomboy [insert pejorative here]. Needless to say, it stopped being that in the next rev, (as I recall), or maybe the one after.
Getting rid of it unfortunately required marking everything (except tomboy) below 'gnome-desktop' as manually installed, and then removing gnome-desktop as well as tomboy, and all of the MONO crap along with it. Or something like that. Ugly ugly hack, but satisfying once complete.
in any case, I think that the devs and "mono fan" maintainers have "learned their lesson" over a decade ago, with the 'tomboy' fiasco. So maybe FUD on the 2nd half too, I guess.
(Even though, Microsoft have successfully shoved a ton of unchecked code into the kernel,
No they didn't. I can assure you that their code contributions have been scrupulously checked (I did some of the checking myself), and nothing gets into the kernel without comprehensive peer review.
For a long time the Windows O/S hasn't been a money spinnner per-say. It's always been things like Office & Exchange. (OEMs were never paying anywhere near full price for their bundled copies of Windows)
With Windows 10 now getting updates/upgrade for free, it does make you wonder what Microsoft's long term plans are for Windows. I feel the obvious answer is that they're interested in the subscription services (e.g. Office 365) over perpetual licenses. There were reports several years ago that MS was pushing its resellers to sell subscription Office 365 over perpetual licenses.
When Microsoft employees turn up at my office with Macbooks & iPhones (soemthing BIll & Steve would never have allowed) it only confirms my suspicions that MS see their future as on-line only. So the thought that Windows may fade into the distance isn't totally radical.
Interesting employees turn up with macbooks, I used to work at an agency who run MS's Australian social for xbox and dabble in other marketing areas and so they worked a day a week onsite to integrate with the client team more, our agency was macbooks (I being one of the few allowed to dualboot with a windows os) but as these guys worked a day a week out the MS office they all got MS gear (mostly surfacebooks) as they weren't allowed Macs, and, to re-iterate, we were a third party, not direct MS employees.
I remember back when I was involved in usergroups and went to lots of Microsoft events... The biggest event I remember was at the Heroes Happen now event that was either for Windows Server 2008 or 2008R2, and, most of the MS evangelists who were presenting had Macbooks.
My time with Windows goes back for 1.0, which was a GUI on top of a separate OS (DOS). So I see irony in the idea that now, after it's multi-step evolution to Win95 and then it's leap the being NT kernel based, Windows might go back to being a GUI on top of a separate (and in this case not controlled by Microsoft) OS kernel and associate bits like device drivers. The "user mode" stuff is to my mind very distinct from the "kernel mode" stuff accessed by calling the kernel's API. What I touch and see is all that matters and Windows is still Windows, so to me a different kernel, Linux or any other, is sorta no difference as long as I get my support in the same seamless way I currently get it form Windows Update which for me "just works".
Running games is the simple part. They have a simple UI and usually don't use the OS widgets and their API anyway. Nor they care about many other aspects a business application need.
That's just show how small is the view of many open source developers - and why most open source software is pretty limited, ugly, and not user-friendly. If people had to pay for it, almost nobody would use it.
Yeah, right.
Games today and historically have always been the activity that taxes the hardware to the max. CPU, GPU and RAM are left choking and sometimes even the disk gets its share (e.g. badly programmed loading screens).
I cannot count the times I have upgraded my PC to the latest and greatest in CPU and video, only to buy a new game and find myself playing at a measly 45fps.
I think that the only other widespread personal activity that can bring a PC to its knees is video editing. I hardly do any of that, but when I do indulge, well, let me just say that I put my gaming on hold until it's done.
Excel doesn't hold a candle to that kind of activity.
"Games today and historically have always been the activity that taxes the hardware to the max. CPU, GPU and RAM are left choking and sometimes even the disk gets its share (e.g. badly programmed loading screens)."
That was not the point of the person you're talking about. They argued that games were easier to emulate than other software, but not because they're low on resource usage. Their argument is that games avoid using a lot of OS-provided libraries. Their examples were the GUI layers, which games often reimplement with their own graphics engines. I haven't ever developed games so I can't comment on whether what they say is true, but please argue against it if you are going to based on the argument made.
"Running games is the simple part."
Bollocks. I decided to install Windows on the unused second drive of my laptop so I could play some games during the lock-down. I've previously been a console gamer, and I was shocked at how much fiddling around with old versions of drivers and third party hacks it takes to get a lot of PC games running in a usable state under Windows. Even on this high spec laptop with Nvidia graphics, I often have to drop the graphics options to very low settings to get non-stuttering gameplay, negating the main alleged benefit of running on a PC rather than a console.
So in summary, running games is hard. Very hard judging by how badly they run on Windows.
>I've previously been a console gamer, and I was shocked at how much fiddling around with old versions of drivers and third party hacks it takes to get a lot of PC games running in a usable state under Windows.
That's because a console only has one specific chipset architecture, making it very simple for games developers. With the PC, games developers are free to specify their own graphics card and other preferences...
It's the reason why my son is only allowed a console, until such time as he can build and maintain his own PC's.
" running games is hard. Very hard judging by how badly they run on Windows."
Yes, I've been discovering this during my WFH Lockdown period, excuse me while I check the 'post anonymously' box, be right back...
Locating and downloading ancient .net dlls and suchlike makes for an interesting and frustrating time.
Having said that, DOS games used to be a bit of a drag as well. I can well remember having to tweak config.sys in order to get sufficient ram below 640k to run Falcon 3. Before the days where the www let you borrow other people's experience quite so easily.
"...and why most open source software is pretty limited, ugly, and not user-friendly."
Gnuplot, image magick, python, r, etc, etc, etc are limited? I think not
I will give you that Windows graphics are stunning and that Windows can be quite easy to use -- when everything goes just exactly right. Sadly, in my experience, that isn't all that often the case.
User friendly? I'll give you that writing bash scripts can be like dealing with a hungry crocodile with a toothache. On the other hand, I don't recall that writing command line scripts for MSDOS/Windows was exactly a transcendent experience. And in my experience Unix tools tend to be far more capable and flexible than their Microsoft equivalents. GUIs? Windows gives you one choice which has devolved from a usable, if mundane, interface in Windows 95 to an incomprehensible jumble in Windows 10. The unix world lets you choose from dozens.
BTW, I'm told that Microsoft has finally gotten around to inventing workspaces -- decades after they have been available in Unix UIs. Really, how have you lived without them?
Seconded here. Though I am not sure that bash scripting is so awful. Maybe it is because I am familiar with it but i prefer it to the OO obsession in Powershell.
Shell scripting should be to get a quick and dirty task done before you put something more elegant in even if that more elegant solution takes decades to happen. ;-)
As someone who has written a LOT of bash over the years (example), I can testify that bash is horrible.
But it's still better than powershell.
Shell scripting should be to get a quick and dirty task done before you put something more elegant in even if that more elegant solution takes decades to happen. ;-)
exactly right!
Whitespace doesn't matter, haven't you heard? (Unless you use some perverse programming language or other, of course. Or Whitespace, which is a fun teaching tool.)
Bloody impressive! --------->
Well thanks, I was hoping someone might like that. :)
(yes, I am that pedantic).
I'm not. At least, not always. As you'll see by looking at my (messy) code. But pull requests are very welcome and pedantry will totally be accepted as valid.
trailing whitespace can at least fool grep jobs not aware of that possibility. Similarly as final line without the new-line character at its end.
OK, to be clear, I said I've written a lot of bash - I never claimed to be an expert: Can you elaborate on what you're talking about, or point me to some relevant resources? Are you just talking about e.g the difference between
^something$
and
^something\s*$
Or am I unaware of something?
Or Whitespace, which is a fun teaching tool
OMG kill me now. This is worse than brainfuck!
Say you recall that somewhere in (many many files) you did set some variable to 'horseapple' but there are many instances of assignments to 'horseapplesomethingelse'. Now you grep for '=horseapple$' but the line really is "BLA=horseapple " (extra blanks at end). No match. This will only rarely byte, but when it does, it tends to hurt a lot. ---->
Btw. by the looks your code is not messy at all.
Aah, right. I don't think I've ever been bitten by this. That particular example wouldn't be likely to bite me because my bash coding style is to always quote variable values, even in simple assignments, i.e:
VAR="somevalue"
rather than
VAR=somevalue
So I would be grepping for "=[\'"]horseapple[\"']" in your example. But I do see what you mean.
Does egrep '\s$' spaceballs.sh find the lines you're talking about? If you care enough to provide line numbers or the appropriate grep incantation I'll be happy to take a look.
by the looks your code is not messy at all
Well thanks. It's not awful - I tend to put some effort into stuff that I release, but it could be better.
Maybe a Tolkien quote is appropriate here: in the foreword to LOTR he says "The most critical reader of all, myself, now finds many defects, minor and major, but being fortunately under no obligation either to review the book or to write it again, he will pass over these in silence".
...aaaand now I feel totally full of hubris, comparing myself to Tolkien. The nerve! ;)
You'll save a microsecond if instead of VAR="somevalue" rather using VAR='somevalue'. That is because strings in double quotes are looked into (so that "${HOME}" would expand, for example). Single quotes means "as it is, do not try anything".
Brought to you from the department of things you never wanted to know.
Argument sentence yours incoherent isn't perhaps compatible. Struggle to understand you I did.
Have never used most open source software you are. Very much not user-friendly, limited, and ugly like say you. Examples including chrome, firefox, libreoffice, and about a million others. OpenGL and DirectX not part of OS API suddenly, news this.
Points good you make. Totally not bridge-living-under creature you are.
> [ ... ] most open source software is pretty limited, ugly, and not user-friendly. If people had to pay for it, almost nobody would use it.
I think a lot of companies pay quite a bit of money for RedHat. [ ?PurpleHat? now that it's part of IBM? ] :-) SuSE and Ubuntu too.
I find myself quite happy with the looks of KDE. I wouldn't call it ugly.
Reminds me of when windows and OS/2 were just emerging.,
Windows was originally to be a different front end to OS/2, one more familiar to people who had experience of Windows 3 on DOS.
But, the story goes that Bill Gates saw the success of Windows and didn't want to share with IBM. So we got the mess that is Windows or was Windows in the past.
As an end user I just want to be able to run the 'Program' I want on my computer whether that is Gimp, MS Flight Sim, iMovie, Office, RaceRender, Forza, Gran Turismo....... etc.
As a Programmer (yes I do write stuff) I want to be able to write in my chosen language (C and its successors) and with just a compiler switch and a few different libraries produce code which runs anywhere.
MacOS runs Linux code with just a few tweaks to makefiles, why shouldn't windows?
"As a Programmer (yes I do write stuff) I want to be able to write in my chosen language (C and its successors) and with just a compiler switch and a few different libraries produce code which runs anywhere."
LOL, oh clearly you're a 733t dev if thats what you think.
Sure , with a few different libraries and LOTS of different code slung between #ifdefs you can write cross platform unix/windows code but there are substantial differences. I'm a unix dev but I have developed on windows now and then and the 2 main differences I came across are:
1) Win32 can't do fork() (though oddly the windows kernel itself can) so you can forget about any sophisticated cross platform multi process code.
2) Windows sockets are not exposed as file descriptors in C (they're a handle which is some structure) so you cannot use select() or poll() to do the network socket multiplexing that is standard with unix and AFAIK you require some over complicated multi threaded or co-routine nonsense to achieve the same result.
3) Unix doesn't differentiate between command line and GUI code ,its all compiles down to a binary which either can or can't connect to an X server. Windows does which brings a load of unnecessary complications IMO.
"But, the story goes that Bill Gates saw the success of Windows and didn't want to share with IBM. So we got the mess that is Windows or was Windows in the past."
More likely that IBM commissioned MSFT to write OS/2 and therefore owned it, which MSFT ended up not liking. Hence the whole subsequent Windows/Windows NT shit show that which was never a decent enough desktop OS until the (painfully insecure, it turned out) XP. All this led to things like Copeland, Taligent and so on when, it turns out, all the world was waiting for was consumer grade equipment strong enough to run a cheap enough UNIX (Linux).
I wonder of Thompson and Ritchie had a clue what they were starting?
It's plausible.
But MS will want Linux running ontop of something to which a licence fee is payable to Microsoft, not a Windows layer ontop of Linux - which would way too optional to many.
It's why we have the Linux subsystem for Windows, the IT answer to building a cathedral on a wet cardboard box.
It's tempting to think if they were intent on switching to a linux base, they'd be working instead on their own Wine-alike - then again - that'd be rather community minded - easier to port their own programs than cater for a platform they were dumping
Re: "But MS will want Linux running ontop of something to which a licence fee is payable to Microsoft, not a Windows layer ontop of Linux - which would way too optional to many.".
I don't think so. Ms appear to have stopped considering os licences to be an important revenue source, with their primary focus being on the xbox, office subscriptions and azure.
I read somewhere that Windows os licences were less than 10% of their income now.
It would not be a Linux desktop, the majority of people using it could not give a stuff about what happens behind the GUI. If it runs the Windows GUI It would still be Windows and most would call it Windows, it is just the hidden bits that would be different.
Most people don't know that the kernel is and for the few that have heard of it, the association is probably nuts and things, not computers.
I am not sure what the benefits for the average user would have to be able to run all their existing software so that needs some sort of emulator. I remain unconvinced that for most people the additional complexity is worthwhile.
So - porting Edge to Linux obviously means that Microsoft is going to move everything to Linux and stop developing Windows.... But they ported it to Mac first. Does that mean they are going to stop developing Windows and move to Mac?
I wonder if the person that came up with this idea stopped and had a bit of a think before he started....
It's the opposite that MS are thinking. After all most of what people do now is in the browser and if the same browser is available on all platforms and it is what people want to use and it is Edge (a bit what if) then why would people not buy cheap Windows PCs?
I think the strategy is clear and far from the prediction made. Windows is the way Microsoft stays relevant to its business clients and OEMs. They're not going to do a bunch of work to get Linux running old Windows code through an emulation layer; it's cheaper to keep supporting Windows, much less likely to break, and it keeps a revenue stream headed their way. Meanwhile, they can't just ignore that legacy code because that's a major reason businesses provide for staying with Windows.
The Edge on Linux part makes total sense to me too. Another major revenue generator for Microsoft is their Office subscriptions, which run through a browser. By porting Edge to Linux, which was probably pretty easy as Chromium already runs on Linux, they make sure that they'll always be able to run Office for Linux users; even if Firefox and normal Chromium make changes that break something MS needs, they can keep it running through Edge. This makes Office more of a cross-platform tool and ensures user satisfaction for not much work.
..Microsoft want control. Chucking a few bones to the OSS world is one thing, ceeding control of their OS kernel is another thing entirely and won't happen. If MS wanted to use a kernel they didn't have to bother to maintain as much they'd do an Apple and use BSD since it doesn't have the GPL hanging over it. The only reason MS contribute to Linux is that there's a monetary payback for them in the form of pulling people from Linux onto Windows via the back door.
"Look, here's our latest greatest on Linux! Like it? Well you can get the full featured version over here --> Whats that? Oh yes, it does say Windows 10 but don't worry about that - look Shiny!"
And Azure and O365 run on Windows Server, and we have some 300 Desktop Apps running on Desktop Windows.
A theory based on MS porting Edge to Linux based on something already cross platform (and they're doing IOS, Mac, Android already with a strong corporate compliance angle baked in) implying a wholesale move is pretty weak.
MS are much more likely to be interested in Linux support for devs + tools for working in the whole IOT space - e.g. Azure Sphere (Linux based IOT). whether that's a Windows box with WSL or a Linux box running O365
This post has been deleted by its author
it would mean that devices for which there are only Windows drivers (Eg strange wifi cards) should cease to be a thing of the past - ie I should have no problems upgrading from MS Windows to Linux when I buy a new laptop.
In reality there will still be some kit (probably from China) that ships with proprietary device drivers for which they refuse to release the source code - completely breaching the GPL and them not caring.
The big problem will always be: tightly integrated GUI or not. The Windows kernel is probably now not that dissimalar from Linux or BSD. But the fun starts with the GUI and multimedia parts for which there is (and there are good reasons for this) still no single, simple system.
But ESR's main point still stands: Microsoft is starting to make a lot of money from its SaaS and PaaS offerings and will be able to keep customised locked into to its APIs for a while yet.
The one thing I will take from this article is the existence of something that allows Steam Windows games to run on Linux.
A ray of sunshine at last.
Now I can go peacefully toward retirement, knowing that, when I get there, I will be able to game to my hearts' content without the clunker that is Windows to hold me back any more.
Happy days.
Last 12 months numbers, roughly
- PC Office, O365, and Dynamics - $50bn and growing 20% pa
- Cloud, server, enterprise - $50bn and growing 30% pa
- Windows, Xbox, Hardware - $55bn and growing 2% pa
So Windows is part of the slowest growing division, which overall is around a third of the business. Where does corporate investment go ? Not in the biggest products, but in the fastest growing - and Windows is a long way off being the biggest anyway. Plus a dollar on O365 is more sought after than a dollar on Windows standalone install, for that lovely lovely subscription renewal.
The business purpose of Windows was to provide a foundation for the real money sales of Office. That is no longer true. Office and Cloud can be sold to any OS user. Why waste money on Windows developers when you can leverage Linux developers who aren't employees and don't qualify for health benefits or a pension.
Upvoted you. Indeed shareholders decide. What is the point investing in an O.S. when you have a free one that works pretty fine. What the corporate user (the one that pays!) wants is to run Office, he couldn't care less of what the O.S. is.
Also devs prefer open source. See all the examples like Hadoop. They are right about that and can see it in the news every day, see at the moment the Apple's tax chronicle. That is what happens when you depend on a company's policy. That can't happen with open source: Libre Office celebrated it's 10th anniversary didn't they.
But the real signals that what Eric describes would be happening is not Edgium. That is too easy since Chromium is already multi platform. The big signals would be:
- Porting Office to Linux
- And most importantly that M$ stop the racketeering business model (forcing down your throat a W$ license with each new PC, whether you intend to use it or not!).
The second signal would really be positive and prove they respect others... but it still makes money, so it's a hard decision!
Until that decision is taken, and as long the racketeering continues (35 years and counting) I'll write M$ and W$, and they won't get the respect of the open source community: where the devs are!
MS Guy 1: Azure? Check. Office365? Check. What the suckers really need now is the EdgeBook.
MS Guy 2: The EdgeBook?
MS Guy 1: Yeah. We can drop all this expensive Windows development crap but keep them mainlining Microsoft while piggybacking on a cheap Linux kernel.
MS Guy 2: Ah, I see. We don't need to make the syringe to keep them wanting the drugs.
Or maybe not. But as I said, I'm not sure.
...but very few of you are remembering the key take-away point from this: ESR is bat-shit crazy. Always has been, always will be. The chances of Linus or Greg allowing MS to Borgify the kernel is about the same as me winning the lottery.
The only grain of concern is people like Poettering who are already in an MS mindset and forgetting the "do one thing, do it well" mantra. Systemd may as well have been an MS designed-by-committee product. It has all the hallmarks. I rebooted a Pi last night with "systemctl reboot" and it felt dirty.
Theonly grain of main concern is people like Poettering ...
There, more akin to how things stand today.
Systemd may as well have been an MS designed-by-committee product. It has all the hallmarks.
Indeed it does.
And I do not discard MS' intervention somewhere.
From a post here at ElReg, May 2017:
"Systemd also has its dirty fingers into other parts of the system. As a replacement for sysvinit is is supposed to be an init system, but because its scope goes far beyond the initialization phase (and it doesn't let you take the good without the bad) it has become a dependency for many userspace programs that should never have any reason to interact with the init system at all, making it harder to use those programs on a non-systemd system."
This systemd is nothing more than a Microsoft registry class virus; the moment I understood what it was all about I stopped using systemd ridden distributions and settled on Devuan Linux.
Now, why would Linux actually need to have something like systemd inside it?
I'll try to sum it up in three short questions:
Q1: Hasn't the Linux philosophy (programs that do one thing and do it well) been a success?
A1: Indeed, in spite of the many init systems out there, it has been a success in stability and OS management. And it can easily be tested and debugged, which is an essential requirement for any OS. Certainly not something you can do with a MS OS.
Q2: So what would Linux need to have the practical equivalent of the registry in Windows for?
A2: So that whatever the registry does in/to Windows can also be done in/to Linux.
Q3: I see. And just who would want that to happen? Makes no sense, it is a huge step backwards.
A3: ....
Cheers,
O.
I'm afraid you misunderstand.
Of course Google/Facebook/Amazon etc. won't use Microsoft's Linux, just like they don't use RedHat or Ubuntu either.
But every corporate, governmental and private user will find it much harder to use anything but the Linux that Microsoft publishes, which unfortunately only works when you pay directly or indirectly a Microsoft tax for its dependence on Azure.
It's precisely the Chromebook/G-Suite, iPhone/AppStore, Android/PlayStore approach, because copycat is what Microsoft has always excelled at.
I wrote an article about a very similar theory a few years ago, no proof though sorry (on an old blog site that got taken down). The idea for the article came from noticing how much virtualisation is already going on with the Xbox One, a platform that obviously cannot afford poor performance. While Microsoft doesn't always make the right decisions around UI, their ability to write seriously impressive system code has never been in doubt in my mind. So it seems highly likely that we will see Windows morph into a Linux kernel, with maybe a couple of virtualisation layers (one for Windows/Xbox games and another for legacy apps), with a Windows-style UI over the top.
I'm convinced that there are people inside of Microsoft looking at what's being spent to maintain the core Windows OS, who are not impressed. That same opinion is probably behind some of the more aggressive marketing in the OS these days.m And with the performance gap that now exists between Windows and Linux, the writing is on the wall.
(Edit: found an archive of the article, a bit out of date now, but just my claim above doesn't sound like BS) https://web.archive.org/web/20180905111053/http://www.blitterandtwisted.com/2017/02/future-of-windows-linux-and-a-built-in-xbox.html)
I've though the same for a lot longer, though I have no archived blog posts to prove anything. ESR is not particularly prescient.
Actually, I think the more useful of your old posts is this one...
https://web.archive.org/web/20180823153529/http://www.blitterandtwisted.com/2016/11/why-windows-will-move-to-a-linux-kernel.html
And, er, yeah, agreed - though I came from a different point of view. The progression to Windows-on-Linux-kernel was clear as soon as the Gates/Ballmer era ended and grown-ups took over. Satya and co watched Google destroy Windows phone with a proprietary user space on top of the Linux kernel. They learned from the pain of Vista and the arm port, and I would personally bet that there's been a functional Windows-on-Linux test rig running in M$ labs for several years now. It'll be very secret but only because the markets might find the idea a bit scary without a *very* carefully crafted announcement, not to guard any particularly secret sauce.
Anyone following RISC-V will note that no RV hardware firms (think NVIDIA) are moaning that a Windows port is needed - unlike the 90s when lack of Windows helped* to take down both alpha and itanium and lumber us with x86 for the next quarter century. RISC-V folk all agree that Linux is vital and if M$ aren't worried, that's because an official Windows userspace will be running on Linux kernels by the time RV64G kit actually hits the streets.
* Yeah, I know there were lots of other reasons, but Not Running Windows was a big one. And no, Windows never worked usefully on alpha
May I suggest an alternative possibility?
First let us recall how MS once tied together IE, IIS, FrontPage and SQL Server with non-standard proprietary knots, thus elbowing out Netscape and HoTMetaL Pro and putting Apache and MySQL in a very hot place.
Now let us spread Edge goodness on to Linux, Linux onto Azure, and build the traditional bait in as proprietary binding - you know, key functionality for the latest MS Office must-haves, the "It may be theoretically (ahem) insecure but it is so productive"s, yadda yadda. Anybody see a pattern emerging?
I emailed Eric Stupid Raymond a couple of decades ago about gun control, just a bland, friendly email. He threatened to shoot me through the internet. He is still yet to develop Bullet over IP.
This guy puts the cray in crazy. He shouldn't be given airtime. Richard Stallman, that's yer guy.
Well I congratulate the tireless efforts of all the open-source developers through the decades.
They kept focus and hacking away whilst 99.9999% of others laughed at them, told them they were not "modern" and told them it would never work.
It just shows that 99.9999% of people really *can* be wrong. If it wasn't for open-source platforms, the industry really would be in a sh*t sorry state right now.
Errm i don't think it has anything to do with the developers. It is more about the fact it was always licenced freely which allowed the curious to play and share and explore different avenues and versions. Open Source development is not always a good thing as there key services that are supported by one developer such as NTP or any of the myriad libraies used with Linux where there is little overview and some really crummy code.
It works well in the case of Chromium, Linux, MySQL and other large Open Source projects because of the support from large companies who provide funding and developer teams (Firefox would not exist without the big corps supporting Mozilla), having multiple teams developiing their own directions and needs and monitoring code of other teams does work, but these are few and far between.
>> Open Source development is not always a good thing as there key
>> services that are supported by one developer such as NTP
True but the same could also be said about closed-source and proprietary development. There is also its fair share of crummy code. The only difference is no-one but the author gets to see it. In your example, if the NTP daemon was owned by a corporation and was closed source, people might be a bit worried in case the owner closes up shop.
I strongly believe open-source will ultimately take us forward. Especially in the "boring" albeit critical things that there is no money to be made in.
The Windows API that so much software has been programmed for. The kernel underneath only matters for stuff like drivers. They can port everything that runs outside the kernel to interface with Linux instead of their kernel, with no performance impact and no changes needed to existing Windows applications. They'd undoubtedly need to develop some kernel drivers to support stuff the Windows kernel handles differently from Linux (probably mostly stuff that interfaces with the Windows Registry) but that's doable.
The side benefit for them is that neither stability or scalability would be an argument for Linux / against Windows on the server side. If they're running the same kernel, none of that matters and Microsoft's advantage over Linux in terms of management/deployment software and a GUI familiar to most everyone would dominate.
Maybe dumping the Windows kernel would make Windows Update a far less painful process, too. Hey, one can dream!
"The kernel underneath only matters for stuff like drivers."
That depends on your definition of kernel. If you don't provide all the IOCTL interfaces and the background services that "applications" use to make stuff happen, you don't have anything useful. It's a vast surface area and while there may be many apps that don't use any of it, and no apps that use much of it, you haven't got a replacement for the Windows kernel unless you have a replacement for all of it. At a guess, it is about 10 times larger than "the kernel". Like the myriad daemons that make a real UNIX system actually useful, most of this code runs in user-space so probably isn't being counted as "kernel". But you need it, all the same.
Of course, underneath all that there's a much smaller/narrower interface that *is* well-defined. If you emulate that, running Windows on top of Linux is easy. But that's not a wild prediction on my part, because MS have already made Windows a reasonably well-behaved virtualisation client.
You treat system calls as part of what needs to be ported. I don't think Windows has an "ioctl" as such, but just as an exchange if the port was in the other direction you'd create an ioctl library call that took the arguments and made calls to the Windows kernel to take the appropriate action, so that application software would run unmodified.
"Assuming this is accurate... Is this how Windows goes ARM?"
No, Windows already went ARM. You can buy a computer with it on if you want. They did this by changing their compiler target to ARM and then fixing broken things until they got tired of that and it seemed to work fine. Linux was not the kernel at that time and it isn't at all proven that it will ever be.
Do I have to explain the difference between CE, RT and Windows x86-64? If I had ment to say the previous Two I would have said so. I said Windows as in Windows X. Making the switch, and pulling it's apps off the x86 Platform al-la Apple.
As to this NOT being Redmonds fist rodeo. Yes we were aware of that. But those products were Windows in name only.
I think this is a good thing and makes sense. Microsoft was always a user experience company, whereas Linux has always been king of the datacenter. Being king of the datacenter means your kernel is faster, more reliable and being FOSS means it is more open and maintainable. So running Windows over Linux actually makes sense. Microsoft isn't stupid, there is more Linux in Azure than Windows now, bur MS continues to make many great products up the stack.
If Azure is based on Windows Server, they would have to write Server with a Linux base to pull this all off and make more money with less development costs. I do not see that happening within the next 5-10 years.
That question was how will Azure change as it is based off Windows Server/Ad not a true LDAP and not Linux.
First off, I think SQL being ported to Linux is far bigger than Edge which had pretty much already been done as part of Chomium.
If Windows were to be subsumed in to Linux as is stated I would see it more likely that Microsoft would do their own Distro of Linux with their own Desktop, as they have done with Chromium where they removed all the Google services and refernces and moved these to MS ones.
They are facing a big problem with their kernal, there is a lot of legacy nonesense in there. Moving to the Linux kernal and wrapping it in a Microsoft desktop with a built in emulation layer for "legacy" Apps in the same way Apple did when it changed from OS9 to OSX, would make some sense.
They have already done this to a lesser extent with Android with the MS Launcher you can get, they are limited to what they can do there by access to the Google Play Store which requires Google services to be installed. Being the dominant/monoploy phone OS and the way the iPhone litgation is going would MS support legal action to remove that requirement? And the an MS distro of Android with MS services wrapped up would be very appealing.
MS is all about selling services now, anything that advances that makes them money, owning the core desktop OS is not a thing for them anymore. Server OS is a little more complicated but needs a reduced feature set and can be a much simpler thing to support, you only have to look at the Nano edition to see where that can go. AD and such like are big things to move as well. But as I have said SQL is already over the fence.
The big thing to watch for would be a Linux version of Office, once that is in place an MS Linux desktop distro would not be far behind. The other biggie would be XBox moving over in the next version due around 2025-7.
This post has been deleted by its author
A Microsoft project for in-car and body-worn camera systems, drones and aerial surveillance systems use the cloud to provide officers with a real-time, connected view of other first responders around them, and real-time situational awareness for officers. Microsoft has partnered with scores of police surveillance vendors who run their products on a “Government Cloud” supplied by the company’s Azure division.
MS has a “Public Safety and Justice” division with ex-law enforcement staff. This is the heart of the company’s police services, andit has operated for years out of public view.
Veritone, along with China-based Hikvision, provide facial recognition services, closed-circuit TV cameras, environmental sensors (for radiation and dangerous chemicals detection), along with automatic license plate readers (ALPRs). In New York over 9,000 NYPD and privately operated camera feed in to the 'beast'. A drone option from Aeryon Labs, the SkyRanger, can provide real-time streaming video.
All these options, not necessarily sold by MS, are plug compatible with MS equipment.
I spent a few months in the US this Spring and a group of us followed the MS MAPP vehicle around gathering data. There are sufficient weakness for disruption, some of which were tested with great success.
Singapore, South Africa and Brazil have systems and Britain will soon join them
If Windows becomes a GUI over Linux. Satya Nadella should win the Nobel prize in Humanity. He will have saved the world from a virus far worse than all others combined, a virus that has infected mankind since that fateful day Don Esteridge, a wonderful man, opened Pandora's box. Don's life was cut short so he was unable to free the world from his fateful folly, but Satya will finally allow him to rest in peace. All that is left is for the god's to grant us one more wish, that Bill Gates somehow does not "Window" healthcare before he enters Dantes eternal buggy upgrade hell. Satya we are forever grateful.
Microsoft has to make sure Windows won't break systems written decades ago.
So the chance of Windows becoming an emulation layer on top of Linux are 0 % imho.
There is no value to Windows otherwise, and Microsoft would be much better off writing a new os from the start.
Also, people have been underestimating the revenues of Windows for a long time.
Windows profits paid for (failed Windows Mobile), for Azure and for Staya's other whishes.
Yes, Azure is becoming increasingly important for Microsoft, but not important enough that it discloses its turnover, much less its profits.
No, it makes perfect sense because YOU have been testing MICROSOFTS translation between kernel APIs for YEARS now. That's what WSL was all about. They took old UNIX based code from NT4 and ported it for this very purpose: to test the shim layer that they are preparing for their move to a new kernel.
but the core Microsoft Executive is likely to be open-sourced for hypervisors to be loaded from firmware like UEFI that runs all operating systems as a guest.. without the wider kernel, the microkernel will revert to the original design.
The "S" in WSL is still a subsystem, but one where the process management is handled within the subsystem and not translated to Win32 calls.
Ain't never gonna happen. Microsoft is going to rewrite Windows on top of a Rust microkernel before it ports Windows onto Linux.
And the entire idea makes no sense. Microsoft can afford to write and maintain its own operating system and has done so for decades. Just because it's cheaper to run Windows on Linux isn't really a thing for a company that makes tens of billions of dollars in profit each year.
No they would not write a new microkernel because that would cost MORE money than maintaining the NT kernel. That manoeuvre would not gain them access to the Linux mindshare, which now dominates many large initiatives. The entire reason for this change is to get closer to the Linux based communities, not further away from them (which a rust microkernel would do). At least Win32 shares the C API!!!
Win32 will become a service that runs inside a KVM module until they can fully port *microsoft* software to the Linux kernel, at which point win32 will become legacy because MS will have the advantage and will go around hovering up more industries in their cloud system.
Interesting Idea however the biggest advantage windows has over linux these days is Active Directory. Yes Samba can do some of it, but it is always multiple releases behind and the tooling is weak. I hope I see it happen one day and I can move off M$ for good. As of now many mission critical apps we use have COM components which is not supported on linux either.
What are your thoughts.
Windows Active directory is a combination of LDAP and Kerberos 5. While the NT implementation of Kerberos 5 is a little bit different from others, these are all userland services that don't specifically rely on a kernel in *nix. It would be trivial to move the code into the kernel to improve performance if needs be.
By the way, I ran a full test domain 5 years ago using Samba 4 and had no issue with most domain level functions I needed.
This has been the writing on the wall for years; I don't even consider development for windows any more, in fact over the past 12 months it has been retired to the occasional virtual machine for legacy software, everything runs on KDE now. Networking with SSHFS is so much better than smb and we have just developed an openssh bridge for printers so that we no longer need VPN and the unnecessary and not insignificant overhead.
Our back office application is written using Ncurses with a portal using web applications written in cgi so our network traffic is now lower than it was way back in 2000 when we were stuck with an ISDN line.
Every single bit of security is now consolidated on openssl so we only have to patch a single source.
It is high time that windows simply became a desktop environment and gave us all a break.
If this happens it would to a degree be a shame. I think it is in Microsoft's best interest to allow it to happen and I am kinda hoping it will happen as well. I just wish it wasn't happening with Linux. Linux is great, for a 1970s bloated monstrosity (slight exaggeration here) but it is still a 1970s bloated monstrosity. I would actually be a bit of a pity of Linux ends up ruling the world. It would be slightly better than if it was Windows, but only slightly.
If the winner was something akin to QNX on the other hand...
i don' t care what the underlying kernel is. An operating system is a piece of code to do housekeeping and provide services to applications ( windowing, gui , storage management , printing spoolers etc ). You don't "work" with an operating system, you work with applications that run on it. Right now the OS is chosen in function of the applications. It would be nice to not have that problem.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
