UNIX's Founders Created Another OS at Bell Labs: 'Plan 9' (bell-labs.com) 135
The team behind UNIX also built another operating system at Bell Labs, writes the corporate CTO and president of Nokia Bell Labs:
Starting in the late 1980s, a group led by Rob Pike and UNIX co-creators Ken Thompson and Dennis Ritchie developed "Plan 9". Their motivation was two-fold: to build an operating system that would fit an increasingly distributed world, and to do so in a clean and elegant manner. The plan was not to build directly on the Unix foundation but to implement a new design from scratch. The result was named Plan 9 from Bell Labs — the name an inside joke inspired by the cult B-movie "Plan 9 from Outer Space."
Plan 9 is built around a radically different model from that of conventional operating systems. The OS is structured as a collection of loosely coupled services, which may be hosted on different machines. Another key concept in its design is that of a per-process name space: services can be mapped on to local names fixed by convention, so that programs using those services need not change if the current services are replaced by others providing the same functionality.
Despite the groundbreaking innovations in Plan 9, the operating system did not take off — at least not enough to justify Bell Labs continued investment in Plan 9 development. But Plan 9's innovations found their way into many commercial OSes: the concept of making OS services available via the file system is now pervasive in Linux; Plan 9's minimalist windowing system design has been replicated many times; the UTF-8 character encoding used universally today in browsers was invented for, and first implemented in, Plan 9; and the design of Plan 9 anticipated today's microservice architectures by more than a decade...!
Starting this week, Plan 9 will have a new home in the space it helped define: cyberspace. We are transferring the copyright in Plan 9 software to the Plan 9 Foundation for all future development, allowing them to carry on the good work that Bell Labs and many other Plan 9 enthusiasts have undertaken over the past couple of decades. Indeed, there is an active community of people who have been working on Plan 9 and who are interested in the future evolution of this groundbreaking operating system. That community is organizing itself bottom-up into the new Plan 9 Foundation, which is making the OS code publicly available under a suitable open-source software license.
We at Nokia and Bell Labs are huge advocates for the power of open-source communities for such pioneering systems that have the potential to benefit the global software development community. Who knows, perhaps Plan 9 will become a part of the emerging distributed cloud infrastructure that will underpin the coming industrial revolution?
Plan 9 is built around a radically different model from that of conventional operating systems. The OS is structured as a collection of loosely coupled services, which may be hosted on different machines. Another key concept in its design is that of a per-process name space: services can be mapped on to local names fixed by convention, so that programs using those services need not change if the current services are replaced by others providing the same functionality.
Despite the groundbreaking innovations in Plan 9, the operating system did not take off — at least not enough to justify Bell Labs continued investment in Plan 9 development. But Plan 9's innovations found their way into many commercial OSes: the concept of making OS services available via the file system is now pervasive in Linux; Plan 9's minimalist windowing system design has been replicated many times; the UTF-8 character encoding used universally today in browsers was invented for, and first implemented in, Plan 9; and the design of Plan 9 anticipated today's microservice architectures by more than a decade...!
Starting this week, Plan 9 will have a new home in the space it helped define: cyberspace. We are transferring the copyright in Plan 9 software to the Plan 9 Foundation for all future development, allowing them to carry on the good work that Bell Labs and many other Plan 9 enthusiasts have undertaken over the past couple of decades. Indeed, there is an active community of people who have been working on Plan 9 and who are interested in the future evolution of this groundbreaking operating system. That community is organizing itself bottom-up into the new Plan 9 Foundation, which is making the OS code publicly available under a suitable open-source software license.
We at Nokia and Bell Labs are huge advocates for the power of open-source communities for such pioneering systems that have the potential to benefit the global software development community. Who knows, perhaps Plan 9 will become a part of the emerging distributed cloud infrastructure that will underpin the coming industrial revolution?
Decades Ago (Score:4, Insightful)
Plan 9 is decades old. Why is this 'news'?
Re:Decades Ago (Score:4, Insightful)
Re: (Score:3)
There are several instances where some old ideas never saw as much as the light of day
Sure, but this isn't one of those instances. Plan 9 is well known to anyone doing OS research and was very influential.
It had many of the same design goals as the Sprite Operating System [wikipedia.org] from UC Berkeley. Neither was widely adopted, but ideas from both are now ubiquitous.
Re:Decades Ago (Score:5, Informative)
Until now plan 9 was proprietary, originally held by AT&T. Nokia, the Finnish company that makes mobile networks, has now made that available as open source.
That is a major deal, if only for history.
Re: (Score:2)
The ideas are important. The implementation is not.
Anyway, Plan 9 has been open-source for more than 20 years.
The "news" here is that the copyright is being transferred to a different organization which may (or may not) change it to a different open-source license. They will matter to no one.
Re: (Score:2)
Bingo. Many years ago I actually installed it and played with it. The youngsters coming online now think they have seen everything. Wrong.
Re: (Score:2)
Re: (Score:2)
I'm pretty sure the Plan 9 that I ran years ago, as a lark, was open-source. And it was frustratingly buggy, but I was trying to use it for weird stuff. Still, I had a few working machines, and dedicated them to the experiment, and it did work. Better for me than any SCO Unix I had to try and figure out.
Re: (Score:2)
Re: Decades Ago (Score:2)
Well, I kept on with Linux, but I was into NetWare back then, so my first working with server was NW 5.1 as a NAMP stack :)
Re: (Score:2)
Netware 5??!! You mean you missed NW 3 and 4? NW 4 was the hot sh1t in its time. The new girl in town everyone wanted to take for a spin around the block.
Re: Decades Ago (Score:2)
I started with NW 2.11. Had a client call me in every few years to upgrade the drives on the DSC they had for a production 2.12 server, eventually up to 1.2GB, which was never contemplated in design. Remember the IDE clock bug? How many buffers could you allocate to handle GBe? Could you use ArcNet and Token-Ring in the same server, at the same time? Why did Novell never fix the Group wise server error redirect bug, when they took years to fix the infamous 'security flaw' they never disclosed? And who the h
Re: (Score:2)
I kicked mars to the curb when I realized it was better to mount SMB volumes right on Slackware or RH, and no one appreciated the NetWare ethos in Linux, not even Novell.
I had a playful uptime contest with two admins I knew well, one san AS/400 manager and the other a RH sysadmin. I won for almost 7 years, and then Novell fixed the clock bug and required a real reboot. The RH SA lost out, and then, the AS/400 manager (my brother) watched an uncommanded IPL execute, from the front panel, looking through the
Re: (Score:2)
That involved buying commercially made CDs by mail (th
Re: (Score:2)
No phone line in either flat. You can't BBS without a telephone line, and the infinite abundance of AOL CDs coming with everything were just laughable.
Actually, you didn't get an AOL CD with everything. The chipship didn't give you one with a pie and chips.
A month-long job in the sands of Arabia paid for both a "large enough" hard drive and enough financial confidence to commit
Re: (Score:2)
Horrible smell when you burn them - not an experiment that was repeated.
Re: (Score:2)
Until now plan 9 was proprietary
What do you mean it was "proprietary"? There was a GPLv2 release of Plan 9 [github.com] back in 2014. And before that it was under the LPL (which is also an Open Source license).
Re: (Score:2)
Re: (Score:2)
Somebody should revive Tanenbaum's Amoeba operating system. It seems to have some similar properties with Plan 9. [cs.vu.nl]
Re: (Score:2)
Seems to be a change of license. It would be like if someone open-sourced OpenGenera [wikipedia.org] despite it's age.
That, and horribly slow / inefficient (Score:2)
Licensing is one issue. It's also horribly inefficient.
It's much slower than Unix and similar.
A distributed operating system is nifty trick. It's not all that useful for day-to-day use. Particularly with 10Mbps networking. It essentially replaced the 128 Mbps ISA bus and the then-new 1,064 Mbps PCI bus with 10Mbps network connections.
Also, the time Plan 9 was actually released, Unix was well-established and Linux was the new hotness.
Re:Decades Ago (Score:4, Insightful)
In March 2021, Nokia Bell Labs transferred the copyright of all Plan 9 software to the Plan 9 Foundation, in order continue the work from Bell Labs for new generations.
https://p9f.org/about.html [p9f.org]
Re: Decades Ago (Score:2)
With the best will in the world no one has given a damn about Plan 9 this century, its a museum curiosity now. This is their way of throwing in the towel without looking like they're doing so.
Re: (Score:1)
We all give a damn!
We're just too lazy to pony up the work effort to melt Linux and an uncompromising implementation of all of Plan 9’s features.
If somebody came along and said he'll pay for that, and present Plan 9's ideas with the clarity of a Richard Feynman, *everyone* in the industry wouldn't be able to write a comment about how much they love it from their *massive* boner blocking their sight.
Re: (Score:2)
It probably SHOULDN'T be. there are a number of good ideas in Plan 9. Some not as good, but that's to be expected in a research OS.
Re: (Score:2)
You express the lack of interest in museum and their contents of the deeply ignorant.
Re: (Score:3, Interesting)
Plan 9 is decades old. Why is this 'news'?
Because there's hundreds of thousands of Linux users who think the Ur-OS is Ubuntu, and installing packages is a skill set on your CV.
Re: (Score:1)
Plan 9 isn't about the past. It's about the future. And remember, my friend, future events such as these will affect YOU in the future [youtube.com].
Re: (Score:3)
Why this is is news is right there in the summary. Granted, the submitter buried the lede, but it's not that long.
Re: (Score:2)
Transferring the operating system to a foundation is the new part.
Re: (Score:2)
I'm baffled too. Perhaps the editors are not actually nerds and have never heard of this before? I actually interviewed at Bell Labs with some of the plan 9 people and saw it in action.
Re: (Score:2)
Re: (Score:3)
Plan 9 is decades old. Why is this 'news'?
Slashdot isn't always the most up to date source of news.
Re: (Score:2)
So true - Plan 9 ran on VMware Workstation almost 20 years ago. You can probably run it today if you find an image.
Re: (Score:2)
Re:Decades Ago (Score:4, Insightful)
When it became available, I switched to BSD on PDP11s and VAXen, and still use OpenBSD (On Sparc64) to this day. I am sure there are Slashdotters older than me, and quite probably with a longer history of using Unix.
I also used (and repeatedly crashed) Multics, and wrote PDP8 assembler on a strait 8.
I was hired to write tape or disk drivers for Plan9 on 68000, but the project was cancelled in about 1985. I believe it was something to do with digital mapping.
My introduction to computers was seeing EDSAC1 running.
Kernels (Score:1)
Re: Kernels (Score:5, Funny)
Python? Thats like so 2010 dude. All the kool kids are totally rocking node.js and javascript! Legit get with the program grandpa!
Just imagine a kernel that loaded parts itself from a random sites on the web at boot with no hash check, what could possible go wrong?
Re: (Score:2)
You mean, Rust crates?
Re: (Score:1)
*Calmly smokes his Haskell pipe, under his CLI tartan, in his Unix rocking chair".
. . . HELP, I'M ON FIRE!
Re: (Score:3)
Are "Rust crates" an actual thing?
If so, please tell me it's because "Rust buckets" was already used for something else...
Re: (Score:2)
Re: (Score:2)
OK. Jr. Calm down ... javascript has been around since 1997.
And yes, I know you are just trolling. But you aren't funny.
Re: (Score:2)
With all the interest in Kernels these days this comes as a welcome surprise. The fight to make the most stable, toughest and robust OS at least in the free-software world is going on. Further the idea of networking with most peers and running a distributed system such as Hurd might not be far way. Plan 9 looks like the idea of Hurd was actually taken up. In a lighter vein I sometimes wonder how it would be if we end up writing the kernel in say Python!! ;)
If anyone was interested in reproducing anything Plan 9 had done, they would have already. Probably decades ago.
This is like Microsoft releasing the DOS source code. Interesting, but not useful.
Re:Kernels (Score:4, Insightful)
Just like we've duplicated everything from Amiga. ;-)
Re: (Score:3)
This is not only not right, it is not even wrong.
Plan 9 was made open source in 2000 with release 3. There have been further releases and licence changes since then.
This is a change of _ownership_, not licence.
Current active forks or continuations of Plan 9 include JehanneOS, Harvey OS, 9front, 9legacy, NIX, Plan B, Arkaros and others.
It is far from dead, and it is more sophisticated and capable than any UNIX, FOSS or otherwise.
Re: (Score:2)
Wait, GNU Hurd claims to have been in development since 1990, not counting the TRIX crap. I don't think there was a release before 2002. I played with Plan 9 well before that. It was running for me some time around 2000, I think, and I was late to it by several years, as my hip compatriots pointed out, though they gave up on it too..
Re: (Score:2)
In KVM/qemu you can allow access to a directory on the host through 9p.
Re: (Score:2)
It nevertheless exists. It has the advantage that the guest is able to freely manipulate ownership and permissions relative to users on the guest system while on the host they are all owned by libvirt.
Re: (Score:2)
The reason Plan 9 didnt take off... (Score:2)
... was because they nailed it first time with Unix. Plan 9 was the difficult 2nd album that ended up a bit meh.
Also FWIW OS services and subsystems in the file system was in unix from the start and was carried over to Plan 9, not the other way around. Eg /dev/tty . However it was expanded by linux over and above what other unixes had with /proc and /sys plus distributed services such as NFS and RPC were made a core part of SunOS back in the 80s.
Re:The reason Plan 9 didnt take off... (Score:4, Informative)
*ba-dum TISS*
Guess you never even looked into what Plan 9 *actually nailed* that UNIX still half-assed (and Linux ruined even more).
Yes, Linux added a few Plan 9 things in. Badly. Like a cargo cult version, missing half the point.
But call me when the network device is part of the file system. Every part of it. No other interfaces needed. And no limitations or slow-downs, but *advantages* and *more efficiency* because of it. (Otherwise it'd also be half-assed by bad programmers.)
Re: (Score:2)
The Subsystems-as-files paradigms only works for things that have a fairly simple interface, hence block and character devices. Network interfaces have potentially unlimited parameters so exposing them as a file makes little sense unless you think having a dozen ioctl's just to create a socket is a sensible idea?
CoC (Score:1)
Re: (Score:2, Insightful)
The Code of Conduct was copied from the seven faculty rules at the University of Woolamaloo:
1. No poofters.
2. No member of the faculty is to maltreat the "Abos" in any way whatsoever—if there's anyone watching.
3. No poofters.
4. I don't want to catch anyone not drinking in their room after lights out.
5. No poofters.
6. There is no rule six.
7. No poofters.
Re: (Score:3)
nice (Score:3, Insightful)
Re: (Score:2)
Have you seen what passes for a nerd these days? Some tween sitting at Facebook with their late programing Ruby on Rails thinks they know all the top shit in the world.
Re: (Score:2)
Re: (Score:2)
Or maybe they explain it as if NOT EVERYONE knows about it.
That is good for a change. Too often the summary uses some concept or acronym that is not common knowledge, and half the readers need to google to figure it out.
Nobody is an expert in all the topics covered here. A good summary will both let the amateurs know the basics, AND let the experts know the interesting details.
I saw the movie (Score:2)
It wasn't very good. And Bela's replacement was a joke.
Re: (Score:1)
sounds a bit like docker (Score:1)
no?
That was a very influencial project (Score:5, Interesting)
It kinda lifted the UNIX philosophy to a new level by making it easy for programs to expose file-system like interfaces. This allows for very well structured interfaces without having to have any special syscalls, libraries or complex parsers. For example there is an IRC client for Plan9. It exposes a directory structure. In its root directory you find a file. If you write a hostname and port into it, it'll connect to that server and create a new directory with that name. This directory will contain another file you can write your channel name into. If you do that you'll get another directory with a file you can write into for writing something to the channel, and another one you can use to read from the channel. Of course all those files and directories are not on disk, but purely reside inside of the program. Since all you need to do is read and write files, it doesn't matter if the program you are using this with was written in a shell script, PHP, C, LISP or Prolog. It's completely language agnostic. This inspired the /proc/ filesystem on Linux for example.
The other famous example is UTF-8 which removes 1-2 layers of complexity compared to other encodings. Essentially it allows you to process UTF-8 strings just the way you did ASCII strings in many occasions while getting rid of code pages some of the weirdness coming with them. 16-Bit Unicode needs really weird encodings as 16-Bit is just not enough to even contain the Chinese and the Japanese character sets. Of course what UTF-8 doesn't solve are the things people are now complaining about which fall back to quirks of natural languages (like i having 2 upper case versions depending on the language) and weird decisions by the Unicode crowd (like encoding flag emoticons by combining several codepoints which encode the ISO code for the country).
Re: That was a very influencial project (Score:2)
Cool metaphors are fun, but that doesnt make them efficient nor particularly useful. That IRC system sounds a pita compared to a standard IRC client interface.
Re utf8 - it doesnt remove complexity, it adds it, fixed size words are much simpler. However it is extremely efficient and compact in comparison.
Re: (Score:2)
In my experience the more metaphors a typical programmer has to understand, the more he'll fall back on a small subset and use them in tortured ways. Everything is a file just cuts to the chase.
Re: (Score:2)
UTF8 is a mixed bag. On one hand, variable length characters, yuck. On the other, legacy programs where an ASCII string is an ASCII string and a NULL byte terminates can handle UTF8 just fine when talking to an appropriate terminal. The other encodings will include "false" null bytes that terminate the string.
Of course with the latest "innovations" in unicode, you can't get away from variable sized encodings. For example, even in utf32, you might have combining codepoints that represent one character positi
Re: (Score:2)
No, UTF-8 existed well before everybody jumped on the 16-bit characters bandwagon. They all ignored it. The few times it did come up it was rejected, partially for the obviously silly (in retrospect) notion that fixed-sized units help (this is obviously useless as words are variable sized, only the most primitive languages limited identifiers to one character). Another reason for rejection was political correctness, they did not want to be seen giving the Americans the "better" 1-byte letters.
Re:That was a very influencial project (Score:5, Interesting)
If we step back far enough and if we’re willing and able to leave some of our inevitable prejudice at the door, we might observe that the design and development of an OS is very much like a perpetual battle of wills. For example, on the one hand you want an OS to be neutral, stable, language-agnostic, extensible and to do an excellent job ‘of the basics’ (such as hardware abstraction, job scheduling, memory management, etc.). On the other hand, we see that the field of general-purpose computing advances most significantly as it goes through the paradigm change - for example the addition of a graphical user interface (something many people today now believe to be inseparable from an operating system).
From time to time we also see attempts at a ‘ground-up rethink’. For example (although I’m not particularly familiar with it) IBM’s OS/400, introduced on the AS/400 servers (now the iSeries, I believe) which was written from scratch as an object-oriented operating system in which everything is an object. Or there was HeliOS, written by Perihelion Software for the Transputer, the 1980’s CPU that was itself designed for the ground up to work in massively parallel implementations (giving the combination of HeliOS and Transputers the means to scale in a fashion not dissimilar to current supercomputers).
Through all this there is a constant tension in OS design - the need for stability and backwards compatibility with the desire to embrace the new. For example, when Microsoft when through the growing pains of switching from MS-DOS6.22 and “Windows for Playgroups” (sic) to their NT (VMS-influenced) platform, they introduced the “Windows on Windows Executive” (wowexec.exe), an abstraction layer that allowed them to run older 16-bit code on a 32-bit OS. (There are practical considerations that need to be considered when asking code to return, say a 16-bit signed integer on a 32-bit operating system (because the 32-bit OS is going to naturally want to return a 32-bit response). If I recall correctly the term used to handle this was ‘thunking’.
It may well be insightful and helpful to look at the best features of Plan 9 and see if they could be integrated in to projects currently under development, but one of the things we have yet to develop is an OS so clean that we’ve been able to maintain a simple, elegant structure, that has evolved with time and which hasn’t become a mess.
Look at the API and you see that Windows is a FrankenOS... It forces oddities on users while allowing developers to disobey rules that results in machines that are unreliable and buggy. Even Windows 10 ‘forgets’ a whole bunch of user-selected choices with each major update. Linux can suffer with similar ills - static versus dynamic linking of libraries can result in multiple copies of common library files on the same machine.
I live in hope of a brave and ambitious group/programmer somewhere taking up the challenge and designing a clean-sheet OS that includes as it’s first order of business a new equivalent to the “Windows on Windows Executive” solution - a sand-boxed environment able to support all legacy code, whilst giving us a clean, brave, new world to explore.
Maybe one day.
Re: (Score:3)
Many people today believe he earth is flat - it does not mean they are well informed.
Hell, some idiots actually use Windows for mission-critical work (at least until the inevitable Ransomware attack).
Re: (Score:3)
But I think we’re each looking at this from a slightly different perspective.
What I’m trying to show is that we don’t have a modern OS - irrespective of lineage - that hasn’t become a rats’ nest of compromises. None of the existing, popular operating systems we
Intentional lack of continuity (Score:5, Interesting)
In 1990, I had a chance to talk with Brian Kernighan about Plan 9, and he made a few interesting points.
First, they recognized at that point that people don't really care where computation goes on. Rather than having a big machine on your desk that you replaced every 18mo, your computation runs wherever it is most appropriate, terminal, file server, CPU server.
To make that happen, he said they wanted to "carry this notion of 'everything looks like a file' much further than it's been carried in the UNIX world." and gave examples of environment variables behaving as files in /env/ and a face server mapped to a directory that gets you someone's picture when you get email from them.
Essentially, Plan 9 was designed as REST services, a few years before the World Wide Web even existed.
The other interesting point was that he felt like some gratuitous changes in Plan 9 were done at too low a level - the system call level - which made it difficult to bring along imperfect, but useful code. But that ties into why the name Plan 9 was chosen. Because the most characteristic thing about the movie is its total lack of continuity.
Re: Intentional lack of continuity (Score:2)
REST is inefficient and evolved as a consequence of the way http connections and web servers originally worked. They're the last thing you want in an OS kernel.
Re: Intentional lack of continuity (Score:5, Insightful)
REST is a religion. Or at least a cult. I've seen it being forced into a tiny battery operated device and when I'd suggest a simple and memory efficient way to do something I'd be told that it was a dumb idea because it wasn't RESTful. REST probably added a couple years onto the project, much of that being meetings where everyone argued about RESTful stuff.
Re: (Score:2)
Re: (Score:2)
I see that I've summoned the Pedantic Patrol.
Obviously, Plan 9 wasn't implemented as REST services. They wanted to but had to change that plan when someone pointed out that HTTP was not scheduled to be invented for several years.
My analogy to REST simply refers to how all interaction gets mapped to a few basic GET/POST operations on a URL space. Plan 9 moves most interactions to a few operations on similar directory and file space.
Except that they were doing it over thirty years ago.
Re: (Score:3)
Congrats on completely missing the point. I wasn't talking about using http in the kernel you plank, I was talking about the stateless paradigm itself. FFS.
Re:Intentional lack of continuity (Score:5, Interesting)
AT&T also had the Blit terminal (5620) where the display computation happened on the terminal but the computation happened elsewhere. This was when you didn't have an underpowered desktop computer on everyone's desk, only a few people actually a computer of their own, and the computer was not called a "server". But the idea was there, predating the X-Windows system, Display Postscript, and such. Download code to the terminal so that you only sent simpler commands to cause stuff to be displayed.
I think this is essentially what you have on the modern web now - except that the modern client display code is inscrutible and in a language with no coherent design and the protocol sends as much data as possible between client and server.
Re: (Score:3)
I Remember (Score:3)
Using Plan 9 at Uni in the early 90s.. Standard UNIX wasn't quite up to the task or Real Time work. Plan 9 was though.
Re: I Remember (Score:2)
Define real time.
Re: (Score:2)
Events as they happened before time travelling was invented. See Svartiblartfasts's efforts in Life, the Universe, and Everything, where he was a leading member of the Campaign for Real Time.
Re: (Score:2)
Re: (Score:2)
That brings me back. H2G2 was made into a popular movie, so the visual reference was easier. Can you imagine someone turning The Cat Who Walks Thru Walls or The Number of the Beast into movies? (And not butchering them.)
Re: (Score:2)
Define real time.
A process is guaranteed the exact resources it requests when it requests them for the duration requested. If you want to know more detail there's quite a few decent papers on the real-time scheduling capabilities of Plan 9 and how it was the first general purpose OS actually able to do this.
Re: (Score:2)
To the microsecond timing comes with consequences - ie other perhaps higher scheduled processes get paged out and the RT process can block everything else if its in a tight loop. You can have proper process scheduling or you can have real time, you can't have both thats why RT linux is a special build.
Re: (Score:2)
Indeed, which is also why these systems have selectable schedulers.
Re: I Remember (Score:4, Insightful)
In a simple manner, if an operation occurs late then it is an incorrect operation. Getting the right answer is not enough, you need the right answer within a well defined time frame. That does not mean just making stuff faster, it means making the timing of operations highly predictable.
Ie, if you need to flip off power within 100ms of a signal being asserted, then that has a real time property. Or at time T you need to switch your radio channel before time T+1ms. Most classical Unix systems have problems here - too easy for one process to block others, paging code in and out of swap makes context switch times highly unpredictable, shoving all tasks through a single kernel will serialize operations meaning a low priority operation can delay a high priority system, and so forth. Also interrupts tend to be non-nested as well. Context switch doesn't necessarily have to be fast for real time, but it does need to be predictable.
Microkernel (Score:2)
It's not true that plan9 is a microkernel or microkernel-like: for instance, the display subsystem (/dev/draw) is part of the kernel. There are also other services which are implemented in the kernel, and even those which are implemented in userland (as some of the actual storage filesystems) could've been moved into kernel without changing the design -- and it's a shame that wasn't actually done, because plan9 was always dogged by very slow disk filesystem and networking i/o, but they chose to stick to the
Did you fall out of the past? Hello? Inferno? (Score:5, Informative)
Isn't there already a successor to this decades old OS, calle "Inferno"?
----
And yes, we really really need "EVERYTHING is a file, and that 9P file system. It would make "the cloud' unneceesary, syncing trivial, and almost anything else trivial too. Like streaming a webcam would just be opening the webcam device file on another computer.
It's ridiculous that people turned to more limited approaches, just because they weren't smart enough to deal with its full power. (Like Linux network devices are said to not be files "because the file interface is too limited". Well, that's why there are directories (records), you idiots! And you can easily extend the file interface with memory mapping, interrupts or whatever you need. It's not the "files". It's you.)
All I would have added, is allow directory hardlinks and file system loops. (You can still crash modern Linux as a regular user, by mounting a FUSE file system that just mirrors another directory ... in a subdirectory of that directory.) ... It goes in *both*. And no, softlinks break, they're crap.)
Because the human brain is not hierarchical. It's a graph. Forcing it to be a tree always results in unacceptable decisions. (Like: Does the "music videos" directory go into "music" or into "videos"?
But hey, we're moving backwards everywhere. Apple products can't hold a candle to a real personal computer, in terms of what you can actually do. They are faster, but what good is it with such a limited toy?
Re: (Score:2)
Are you drunk again? Very little of what you're writing makes sense. Is the "unnecessary cloud" you're talking about that magic smoke that is floating around your apartment while you envisage everything being a file while listening to Pink Floyd and marvelling at the colours in your Plan 9 terminal?
Re: (Score:2)
Late by two decades, if not more ... (Score:2)
This action by Bell Labs would have been very useful if it happened in the late 1990s or so, before Linux replaced UNIX.
Oh, and there is already Lucent Inferno [wikipedia.org], which was borrowed a lot from Plan 9.
Looking at the history, it has been open source since 2004, which I don't remember last I checked, many years ago ... And it is no longer owned by Lucent.
It's so good to see Glenda again. (Score:2)
https://p9f.org/glenda.html
Re: (Score:2)
I'm never sure of it's name, it's either Glen Or Glenda.
Plan 9 has been open source since 2000 (Score:2)
Lucent lost interest in Plan 9 and just used a few concepts in Infero, a distrbuted OS designed to compete against Sun's Java Platform, which also failed and was abandoned.
Kernigan, Pike et al. continued to use Plan 9 and play around with it. They were allowed to release the 3rd Ed in 2000 and 4th Ed. 2002 under incompatible open source licences. The last offical release was 2015. The reason for this post is that ownership has been officially transferred from Lucent to a Plan 9 Foundation, who have put ev
Ah, Plan 9. (Score:2)
A decent idea. Inferno was the spinoff. I have the CD for that somewhere. I've seen lots of complaints about code quality, though.
I'm not sure if it'd fit into the definition of nanokernel, where practically everything is a servce (an extreme extension of the microkernel idea), and the concept should certainly be kept going.
But going back to Plan 9, rather than Inferno, seems odd.
WORM keylogging feature (Score:2)
I haven't used it but always thought Plan 9 had an amazing capability of recording all of your keystrokes on a WORM drive so that you never lost anything and could always go back in time. My mac's time machine is not as reliable and I always wished I had that. Even just this week I experienced an editing session this week with too many things in my head, mistyped something and had to go back and forth with git diffs to figure out what I had done. In word processors you only have one linear forward and backw
Dupe! (Score:2)
Plan9 is now Officially Open Source [slashdot.org]
"Posted by CmdrTaco on 10:22 AM June 17th, 2003 from the now-isn't-that-special dept."
Re: (Score:2)
That's because EditorDavid is a basically moron. Check other topics he curates and you will fund he lacks a basic understanding of the history of computing.