Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror

Driving Plan 9 226

Posted by CmdrTaco
from the because-you-can dept.
Glenda_lives_on writes "OSnews has an alternative OS review on Plan 9. Plan 9 is a research OS produced by Bell Labs. It was open sourced a few years back, and has enjoyed a revival of sorts. Los Alamos National Labs is continuing to favor Plan 9 for their new generation of super computing because its the fastest thing out there. I have downloaded and ran Plan 9 before. In fact the Plan 9 live cd sits here on my desk. Its not an operating system for noobs however, and lacks some graphical refinement. Plan 9 is a very cool and a interesting test drive however. Its definitely worth the price of admission (free) for exploring, and education."
This discussion has been archived. No new comments can be posted.

Driving Plan 9

Comments Filter:
  • by writermike (57327) on Sunday July 23, 2006 @09:20AM (#15765682)
    Were Plans 1-8 "not entirely successful?"

    "You see! You see! Your stupid minds! STUPID! STUPID!"
    • Now YOU look stupid. (Score:4, Informative)

      by Anonymous Coward on Sunday July 23, 2006 @09:26AM (#15765695)
      Plan 9 is a reference to the Ed Wood movie, Plan 9 from Outer Space, often regarded as the worst movie of all time. Aliens raise the dead to finally prove to humans that they exist (because that's certainly the most obvious, effective way to do it).
      • by Anonymous Coward on Sunday July 23, 2006 @10:00AM (#15765771)
        Considering that the "Your stupid minds! STUPID! STUPID!" line is a quote from the movie, I suspect the OP knew that.

      • by andreMA (643885)
        Actually I took the "not entirely successful" to be a reference to the TOS episode The Ultimate Computer, where (when discussing the M-5) Kirk asks Daystrom (the inventor) about M-1 through M-4. Daystrom responds that they were "not entirely successful"

        I thought it was funny, playing one of the cheesier TOS episodes against the extremely cheesy Plan 9 From Outer Space

      • by gkhan1 (886823)

        Ahh, but you are mistaken in your praise of Plan 9, most of us Ed Wood aficionados knows that that isn't his masterpiece, indeed it is trivial in comparison to the behemoth of Glen or Glenda [imdb.com] (also known on sensational posters as "I Changed My Sex!"), wherein Wood himself stars as transvestite and Bela Lugosi is an insane rambling doctor.

        All jokes aside, Bela Lugosi really deserved better than Ed Wood. It's a shame to see this man who scared the living daylights out of so many people with his Dracula and re

        • Unfair to Ed Woods (Score:4, Interesting)

          by Doug Merritt (3550) <`gro.euqramer' `ta' `guod'> on Sunday July 23, 2006 @10:52PM (#15767675) Homepage Journal
          Bela Lugosi really deserved better than Ed Wood. It's a shame to see this man who scared the living daylights out of so many people with his Dracula and really made a mark on movie history be reduced to lap-dog in the hands of a complete hack. I guess Wood helped him make another mark on movie history.

          In some sense, sure, but on the other hand, that is grossly unfair to Ed Wood's beneficial relationship to Lugosi. Consider the following (from http://en.wikipedia.org/wiki/Bela_Lugosi [wikipedia.org], but which agrees with similar details from some films on the subject, including one with Johnny Depp as Ed Wood, from some years back):

          Later on, the acting jobs dried up, and Lugosi became addicted to morphine...Late in his life, he again received star billing in movies when filmmaker Edward D. Wood, Jr., a fan of Lugosi, found him living in obscurity and near-poverty and offered him roles in his films, such as Glen or Glenda...in Bride of the Monster...and in Plan 9 From Outer Space [*]

          ( [*] Plan 9 was a posthumous performance, and note that the death of Lugosi early in filming was precisely the reason that Plan 9 became "the worst film ever made" rather than merely the usual Ed Wood grade B movie.)

          Yes, Wood was a hack (albeit a fun one with a cult following to this day), but he did his best to rescue Lugosi when the rest of the world had given up and no longer cared. Give credit where credit is due, rather than simply sneering at the charitable, no matter the flaws you see in the good samaritan. By all accounts, Wood seems to have done the best he could by Lugosi.

    • If they had used letters, I wonder how many times the developers would have purposely bombed the project just so they could get away with calling it "Preparation H"...
    • Were Plans 1-8 "not entirely successful?
      It could be worse. I still wonder about the poor bastards who were in the clinical trials for Preparations A-G.
    • While there's a clear Ed Wood movie reference, the reason it's something-9 is that it was the Next Thing after Version 8 Unix, which was the Bell Labs release after Version 7 Unix, the one that Started It All. So clearly, rather than call it Version 9 Unix, when it wasn't really Unix underneath anymore, etc. they needed a new name...
  • Surely... (Score:2, Funny)

    by isecore (132059)
    ... Ed Wood must be proud :)
  • Plan 9 ISO Mirrors (Score:5, Informative)

    by ettlz (639203) on Sunday July 23, 2006 @09:26AM (#15765691) Journal
    Before y'all go pulling down the ISO to try it out, the mirrors are listed at http://netlib.bell-labs.com/wiki/plan9/Mirrors/ind ex.html [bell-labs.com] .
  • by jpellino (202698) on Sunday July 23, 2006 @09:27AM (#15765697)
    ... are cheesy 10 ft tall silver curtains.
  • Plan 9 is cool (Score:5, Informative)

    by Anonymous Coward on Sunday July 23, 2006 @09:28AM (#15765702)
    I've never used it, but Plan 9 offers a radically different archival storage system called Venti [bell-labs.com].

    Basically it never deletes old blocks of data from the server. Blocks are write-once, identified by a really large hash (collisions are so improbable that the possibility can be totally ignored). This allows you to copy lots of redundant data to the server (such as periodic backups) without worrying about the storage space. If the blocks were ever copied there before and they have not changed, they won't take up any space!
    • Sounds a bit like ZFS.
      Never mind which was first, ZFS is available today.
      • Re:Plan 9 is cool (Score:5, Informative)

        by ciroknight (601098) on Sunday July 23, 2006 @10:42AM (#15765849)
        Both are available today, and I can assure you Plan 9's Venti was first, and furthermore ZFS isn't really anything like it. Venti does data compression by removal of redundant data by (basically) writing a block, and then checksumming the block and using that sum to refer to the block in the future, so that if the software tries to write an identical block, it simply ignores the request. With an appropriate block size set, it can save lots and lots of space, however, it's very impractical as a day-to-day filesystem due to the datasets most people work with day-to-day (most of us work with lots of non-redundant data such as code files, video files, image files, etc.), though it would be a neat experiment to see what could be done with a modernized version of it.

        Lots of things like this were/are revolutionary about Plan 9, simply because they were given the ability to do it. Some of them are great ideas (like Venti in conjunction with a database server, if the database server was tailored to the file system and didn't do stupid things...), and some of them could still use a great deal of work. Either way, I welcome our Plan 9 overlords from Outer Space.
        • Venti does data compression by removal of redundant data by (basically) writing a block, and then checksumming the block and using that sum to refer to the block in the future, so that if the software tries to write an identical block, it simply ignores the request.

          That sounds like a hell of a lot of overhead. With so much stuff going on, what do the performance benchmarks of this fs look like? Doesn't it pretty much guarantee fragmentation? I mean, if you have a lot of redundant data there is no way it
    • How much is this......Venti? Is it MUCH more expensive than a Tall?
      • by hey! (33014)
        No, but the CDs are roasted way too dark. Of course some say that Americans are just too parochial to appreciate properly carmelized bits, but of course if it's so much better that way, why does their port of mkisofs have the --with-mocha and --with-vanilla-syrup options?
  • by DrSkwid (118965) on Sunday July 23, 2006 @09:30AM (#15765709) Homepage Journal
    man page != implementation

    http://cm.bell-labs.com/magic/man2html/1/emacs [bell-labs.com]

    NAME
            emacs - editor macros

    SYNOPSIS
            emacs [ options ]

    DESCRIPTION
            This page intentionally left blank.

    SOURCE
            MIT

    SEE ALSO
            sam(1), vi(1)

    BUGS
            Yes.
    Copyright © 2006 Lucent Technologies. All rights reserved.

    and vi(1) [bell-labs.com] isn't what you might think either
    • Yeah, I couldn't believe TFAuthor wrote that. I fired up my Plan 9 box just to make sure it wasn't true.

      The vi thing I found a long time ago; at first I thought it was to prevent anyone from porting vi to their system but since realized that it was just a result of the naming convention for their development tools.
  • Oh, slashdot (Score:5, Insightful)

    by Zouden (232738) on Sunday July 23, 2006 @09:31AM (#15765713)
    I have downloaded and ran Plan 9 before. In fact the Plan 9 live cd sits here on my desk. Its not an operating system for noobs however

    What is this, digg?
  • Zzzzzzz..... (Score:5, Interesting)

    by countach (534280) on Sunday July 23, 2006 @09:39AM (#15765730)
    The "everything is a file" metaphor of Unix was revolutionary at the time, and Plan 9 taking it a little further really does little to advance the state of the art.

    What was good about the "everything is a file" metaphor was not the "file" part, but the "everything is a...." part.

    What would really advance the state of the art is an "everything is an object" operating system. It would be something like a Lisp OS but with an object database type file system. I think some have existed in academia, but I've never looked into them.

    • Re:Zzzzzzz..... (Score:4, Insightful)

      by ems2 (976335) on Sunday July 23, 2006 @09:43AM (#15765739) Homepage
      "We have persistent objects, they're called files." -- Ken Thompson [bell-labs.com]
      • "We have persistent objects, they're called files." -- Ken Thompson


        Hmm, how does one subclass a file to customize its behaviour? Or hide its implementation details by marking them private? Or call methods on it?

        • Directories? chmod? open()?
          It worked for udev/sysfs/procfs.
          • It worked for udev/sysfs/procfs


            Well, if by "worked" you mean they were able to extend the kernel to do clever things and have the results look like files, then sure. But in something that is meant to be functionally equivalent to persistent objects, I would expect to be able to do all of the above from any userland application.

      • Except that normal objects have a complex interface (public methods) whereas files are restricted to read/write interface.
        So these two things are very different..
        • You may see this as a shortcoming but it's really not. What if every file was actually an object? You would have to keep track of what each object was, you would have to catalog the public interface of every file, you would have to rewrite every single application everytime a new class was created so it could deal with the new objects.

          Programs like grep, sed, vi, etc would instantly become useless because they would have to know about every single class in the system, past present and future!.

          It would be a
          • 'A shortcoming' imply a value judgement, I just said that files having a very rigid interface are not true objects, that's all.

            Just a remark: the read/write interface is well suited to console/script interaction, but now applications have mostly moved to GUI, where the 'everything is a file' is not so useful: as evidenced by the fact that Plan9 didn't manage to fit a 'mozilla' like web browser in their setup.

            Also, the textual interface while being very useful has also some inherent difficulties: many script
          • Just pulling stuff out of my ass here, but...

            The superclass, File say, would have all the methods that we normally associate with files. That is, it would represent the file as a stream of bytes that you can read from or write to. Since many files are plain text files anyway, all the normal ways of processing text apply.

            Then you could have the ImageFile subclass, for example, which supports a method "Pixmap ImageFile::RenderImage ()". If the OS has the right libraries installed, it will abstract away fr

            • In that case you would be exactly where you are now. You could do common things like read lines, read bytes, write lines, write bytes, check for EOF etc but for anything more sophisticated you would need some code that knew about the file. No different then what you have now. Right now I could serialize and object written in any language and save it to a file. In order to do anything with that file other then read bytes or lines I would have to have the original language and perhaps the serialization routin
          • If you really want to interface with things via a crappy grep/sed interface, you can always do a toString() on an object. This will allow you to retain all the bugs that traditional shell script sed/grep parsing programs have always have. But if you want to move up in the world you could use a useful interface to the object.

    • The all-encompassing namespace isn't a new thing, no, viz. UNIX and NT. I think Plan 9's revolutionary idea is taking the "everything is a file" idea and distributing it (devices, files, everything) across many physical machines. Does even VMS clustering have this ability?

      As for the holy-grail object systems (with all the bellen and whistlen such as orthogonal persistence) there were projects like EROS, Coyotos, etc., and some early projects like DERA's Ten15.

    • Re:Zzzzzzz..... (Score:2, Informative)

      by Bobjects (261759)
      What would really advance the state of the art is an "everything is an object" operating system.

      Smalltalk-80 fits that description pretty well. See:
      http://users.ipa.net/~dwighth/smalltalk/byte_aug81 /design_principles_behind_smalltalk.html [ipa.net]

      Smalltalk-80's modern descendent is Squeak:
      http://www.squeak.org/ [squeak.org]
    • Re:Zzzzzzz..... (Score:5, Informative)

      by ratatask (905257) on Sunday July 23, 2006 @11:46AM (#15766023)
      >The "everything is a file" metaphor of Unix was revolutionary at the time, and Plan 9 taking it a little further really does little to advance the state of the art.

      Sorry, this is where you're wrong.
      Make the socket interface a filesystem, and all you do is mount a fileserver over that, to create
      a socks proxy/http proxy/whatever. All apps get the capability of
      doing networking over a proxy, transparently - no need for using libs or
      prelinking hacks that usually don't work.

      Have the ability to easily create fileservers in userspace, and create an mail
      filesystem that can handle imap/pop/local mboxes etc. Mail clients doesn't need
      to reimplement your favorite mail protocol in yet another broken and incompatible
      way, or adhere to 4 different libraries with 4 different concepts. Just read/write files and
      have the one fileserver do the job.

      Sharing files AND resources becomes easy too. Want to play sound on another computer ? import hostname:/dev/audio /dev

      Having all resources being files, you get a standard way of access control (add ACLs if you really need to), couple it with private
      namespaces, and you don't need the umpten hacks like freebsd jails, chroots, selinux, systrace, etc. Just use chmod/chown and set up a filesystem namespace only containing the resources (resources in this case is anything you request from the OS - networking interface, audio device, screen display, authentication privileges, or most other of the 400 syscalls or ioctls you might want to restrict access to in a read/change on traditional unixes.

      • Having all resources being files, you get a standard way of access control (add ACLs if you really need to), couple it with private namespaces, and you don't need the umpten hacks like freebsd jails, chroots, selinux, systrace, etc. Just use chmod/chown and set up a filesystem namespace only containing the resources (resources in this case is anything you request from the OS - networking interface, audio device, screen display, authentication privileges, or most other of the 400 syscalls or ioctls you might
    • What would really advance the state of the art is an "everything is an object" operating system.

      What's cool about "everything is a ..." is that code and understanding for dealing with one entity (such as disk files) can be applied toward other entities (such as network connections). "Everything is an object" only helps if the objects share important characteristics that make them similar to work with. Using a basic definition of "object" -- a piece of data and operations for manipulating it -- Unix qual

    • Actually Plan 9 does far better than "everything is an object". In Plan 9 everthing is a subclass of a certain base object, called a "file". Programs can assumme a set of useful functions, such as the ability to copy an object, that actually makes this api useful, rather than an academic experiment.

      About all that "everything is an object" means is that it wont crash if you send the id for one type of object to an api designed for another type, instead you will get an error. Really the only difference betwee
    • What would really advance the state of the art is an "everything is an object" operating system.

      Why is a file not an "object"?

      I don't get this. We call them 'files' and try to think of them as pieces of paper and folders as boxes, but the things on out harddrives are very different from their real-world analogs. They are abstract collections of data.

      There are well known methods for getting information about a file: name, contents, attributes. There are well known methods forming an orthogonal action set u
    • NT does have an "everything is an object" metaphor, at least for things exported from kernel to user mode. The Object Manager defines a class for such objects, of which object types are subclasses. Kernel mode components are free to define their own object types. The Object Manager defines the interface for common object functions and the owner of each specific object type defines an interface for specific functions (usually by way of syscalls).

      Each object type has a set of function pointers for optional m
    • There was an operating system back in the 80s called 'Walnut' which had everything as persistent objects with various economy things like rent and capabilities. If you (or another process) didn't pay an object's rent, the object would die. It was quite interesting though not really developed to the point that people could use it.
  • Considering its not really intented as a desktop replacement OS, thats by design.
  • by ems2 (976335) on Sunday July 23, 2006 @09:59AM (#15765768) Homepage
    The review is not so great in terms of accuracy i.e. there is no emacs (check out acme, sam, ed, and smacme instead) and the 640x480 resolution is nonsense. 9fans [psu.edu] certainly isn't so grateful about this review. [psu.edu]

    Check out the Plan 9 documentation [bell-labs.com] if you are interested in understanding Plan 9.
    • Hi Ems2, I am the author of the referenced Plan 9 article and I can safely say that 640X480x8 was the default window size that came up.This is fact and not opinion.I think I mentioned that it was easily changed to a higher resolution, either by Rio or by simply typing 800x600x8. I now see that there is a version of emacs available. I referenced an out-of-date posting on the 9fans list (I think). Thanks for the correction. Emacs is a great asset. You will see that I too provided plenty of references to th
    • 640x480 should be good enough for anyone.
  • by DrSkwid (118965) on Sunday July 23, 2006 @10:00AM (#15765770) Homepage Journal
    "Plan 9 from Bell Labs" is the proper name of the OS.

    Plan 9 is now community driven, albiet from a small community, mostly the same people that have been there all along.

    It has USB sound support and AC97 support is a new one on me.

    I use it still because the user environment is the best one I have encountered for text editing and interecting with the shell. Most users use VNC to get to their X11/Windows desktops where their web browser lives.

    Building a web browser from scratch is one of those never ending tasks that frankly, just isn't worth your while. That said there is Mothra - no tables, no css, no frames etc. that Tom Duff (yes that Tom Duff) wrote many moons ago and one of the community is beavering away at his project Abaco and has moderate success.

    One of the main tenets to Plan 9 is "everything is a file" and the system is built around the notion of a distributed name space in the shape of a directory tree rather than being a reflection of the disk contents. The canonical example of this is ftps where the remote ftp site is presented as a directory tree at /n/ftp

    Name spaces are process independent so you can build them per process which feels a bit like chrooting.

    Exporting a name space is part of the deal, this presents many gifts that were not deliberately shoe-horned in such as remote step debugging across architectures, sending sound to a remote soundcard, importing a remote machine's network stack instead of using a gateway (including non-plan9 machines via ssh), importing remote filesystems (including non plan 9 machines). All this is facilitated by the 9p protocol [bell-labs.com].

    As a micro/macro kernel hybrid all this is achieved in just 37 syscalls which is a source of amusement and a feeling of superiority when compared to Linux' 300+ (so many they are not even enumerated any more).

    Linux is derided in the mailing list ("For amateurs, by amateurs") as well as the failings of the other braindead OSes we have to deal with ("If only they did it like us").

    Linus has stopped by in 9fans to whine on about stuff and was seen off, Theo wanted our compilers when he didn't want the license (as imposed by Lucent lawyers) but since they have been dual licensed we've not seen him around.

    Inferno isn't plan 9, it's another product built on similar principles that was sold off by Lucent.

    Lucent's management of Plan 9 in hindsight could have prevented adoption when it was crucial - it was $300 per copy prior to v. 3 and once a free download had a "copies of all modifications must be sent to Lucent" clause and other annoying restrictions in it. These have been lifted now but they boat could already have sailed.

    The notion of distributed computing has gained ground in recent times and Plan 9 could have been at the forefront with distributed computing being built in from the start.

    All that said, Plan 9 was never intended as anything more than an experiment and some ideas have slowly crept into other products (or possibly independently invented) - notably Windows XP presenting their stuff as files/folders, ftpfs in Linux, single sign-on.


    • Exporting a name space is part of the deal, this presents many gifts that were not deliberately shoe-horned in such as remote step debugging across architectures, sending sound to a remote soundcard, importing a remote machine's network stack instead of using a gateway (including non-plan9 machines via ssh), importing remote filesystems (including non plan 9 machines). All this is facilitated by the 9p protocol [bell-labs.com].

      This sounds remarkably similar to what Richard Stallman's The Hurd was suppose
    • > Theo wanted our compilers when he didn't want the license
      > (as imposed by Lucent lawyers) but since they have been dual
      > licensed we've not seen him around.

      Is there any license for the compilers other than
      http://cm.bell-labs.com/plan9/license.html [bell-labs.com] ?

      It sure looks it's written by lawyers, though earlier versions was worse ;-)
      • You can also get Plan 9's c compiler (by Ken Thompson) from Inferno [vitanuova.com]. FYI, they are kept in sync. The only difference is the license used with the version bundled with Inferno. The license is based on MIT-template. The text goes as follows:

        This copyright NOTICE applies to all files in this directory and subdirectories, unless another copyright notice appears in a given file or subdirectory. If you take substantial code from this software to use in other programs, you must somehow include with it an appropr

    • So, basically you're saying that you have the most impressive OS on the planet and the rest of us are simply idiots for not recognizing just how wonderful it is? Or are you saying that you have the best OS on the planet and your /glad/ that the rest of us haven't recognized that fact so we can't spoil your tiny playground?

      Or is it that you hate the fact that an OS built "by amateurs, for amateurs" is somehow eating your lunch at every turn? Are you happy living with an addmittedly experimental OS that doe
      • If it were simply a matter of being a better OS, OpenBSD would have flattened Linux a long time ago.

        I think you mean: if it were simply a matter of having an idea how to build a better OS, OpenBSD would have flattened Linux a long time ago. The big problem with OpenBSD is that they haven't built most of it; it's sorely lacking in (native) features. It doesn't matter how good or bad your kernel is, emacs and firefox are basically the same on every platform. OpenBSD has so few actual (working) pieces that it'

        • I think you just proved my point for me: If Theo and company would have been more willing to accept contributions gracefully at the beginning of the OpenBSD project, I contend that many (most? all?) of those problems would have been dealt with a long time ago. Instead, Theo seems driven to make sure that only his vision will prevail, and he doesn't hesitate to slag someone who he disagrees with. He doesn't seem to have learned much about how to play nice with others after being booted from the NetBSD gro
      • blah blah blah

        Your assumption that I quote myself is erroneous.

        But yes, I don't care if you use it or not.

        Worse is better worked out for Linux & HTTP/HTML

        Market share is for marketing people.

        Not everyone is trying to dominate the desktop.

        • Worse is better worked out for Linux & HTTP/HTML

          Market share is for marketing people.

          Not everyone is trying to dominate the desktop.

          Hmmm. Interesting. And yet, you care enough to poke everyone in the eye with how lousy you think Linux is that you have to post this for a sig:

          GNU/Linux - a printer driver gone horribly wrong

          You know, for someone who claims not to care about what the rest of us do or think (i.e. who has the bigger marketshare), you sure seem to have spent a lot of energy th

    • I use it still because the user environment is the best one I have encountered for text editing and interecting with the shell. Most users use VNC to get to their X11/Windows desktops where their web browser lives.

      "Instead of driving a car, I get about using a pogo stick. It's much better because you can carry it with you when you're not using it, so you always have it available, and also you can get into smaller spaces than you could with a car. When I need to go someplace that's too far away, I use my car

    • It's really easy to make operating systems elegant if you aren't constrained by the terrible task of supporting a large base of (often old) software and hardware (particularly software).

      Nothing epitomizes, to me, the Plan 9 attitude like your remark later in the thread about how not having shared libraries is something to celebrate (it's a feature, not a bug!). I'm aware that you didn't just make this up; it's a pretty common thing for Plan 9 implementers and fans to say. Yes, shared libraries are awkward t
  • by ems2 (976335) on Sunday July 23, 2006 @10:54AM (#15765886) Homepage
    First of all there is Charon [vitanuova.com] from Inferno [vitanuova.com]. It supports html, EMCAScript (1.1 IIRC), CSS, DOM (level 1 IIRC) and https. (See screenshot [operating-system.org] however this one is a bit outdated)

    Abaco [freshmeat.net] is the most actively developed Plan 9 web browser. It supports most of html. DOM level 3 development has been started. Mozilla's Javascript engine has been ported to Plan 9 and can be used today for a Javascript shell. This will provide abaco with Javascript in the future. Work on CSS has started but I do not know what has been done or where it is heading. Abaco has been ported to Linux and friends via Plan 9 from Userspace [swtch.com]. Package managers are encouraged to make packages of abaco for their systems. (See screenshot [tip9ug.jp])

    Then there are webpage, links [ucalgary.ca], mothra [wikipedia.org], and htmlfmt [bell-labs.com].

    Finally there are text web browsers for acme (htmlfmt for Plan 9 and see this [caerwyn.com] for Inferno)

    In other news, SDL now works on Plan 9 [tip9ug.jp].
  • by ChePibe (882378) on Sunday July 23, 2006 @11:16AM (#15765937)
    Plan 9 is now available for free from Google Video [google.com].

    I must warn you, however, that everything you will see is based on sworn testimony...
  • /proc on steroids (Score:5, Interesting)

    by DrDitto (962751) on Sunday July 23, 2006 @11:19AM (#15765949)
    My university had a visitor from Los Alamos several weeks ago and he gave a live demo of using Plan9 to control a 10,000-machine cluster.

    Really cool how _everything_ was a file.

    To start a program on some machine, he would cd to some directory corresponding to the machine. I don't remember exactly, but this directory had files corresponding to "exe", "stdin", and "stdout" among others. To start a job, the program was just copied to the exe file. And then if you looked at the "stdout" file, the output from the running job was there. Now you can imagine how launching a job on thousands of machines and collecting the output becomes really trivial.

    I got the impression that this was sort of like the Linux /proc filesystem, but expanded to work seamlessly across a cluster and with more functionality.

    • Re:/proc on steroids (Score:2, Informative)

      by ems2 (976335)
      Linux got the idea for /proc from Plan 9 [bell-labs.com]. However, it a very dumbed down version of Plan 9's. One of the major differences is that Plan 9's /proc controls processes while Linux really does nothing but represent them to some degree. One example is that you either kill a process by writing 'kill' to its clt (control) file or delete its directory. Plan 9 requires less syscalls thanks to this design. Inferno [vitanuova.com] also has this design to manage its processes. Imagine this with Plan 9's distributing ideas...
    • Re:/proc on steroids (Score:5, Informative)

      by spitzak (4019) on Sunday July 23, 2006 @12:36PM (#15766163) Homepage
      Linux copied the idea of /proc *from* Plan9, so there is certainly some resemblance!

      Despite the limited nature of the copy (somebody else says that Linux version is pretty much read-only, I'm not sure) it shoud be obvious what a big win this is. Suddenly a whole lot of utilities like "ps" do not have to be recompiled to match the kernel. And you can peek into /proc directly, without using a program, and get useful information (such as what files are open or the executable name, I've done both of these plenty of times).

      The only other Plan-9 thing that is copied extensively is UTF-8 text encoding. This one is also a HUGE win, as suddendly we don't have to write two streams through all our programs for handiling Unicode and handling "legacy" ascii files, as they are now the same thing, as long as some (very minor) fixes are done to the "legacy" code. Plus UTF-8 seamlessly handled Unicode going past 65536 characters, while the "wide character" solution that Sun and HP and Dec and Microsoft struggled with for 20 years fell apart the moment this happened, by adding "surrogate characters" and thus deleting the *only* advantage it had over UTF-8.

      Considering how incredibly useful both of these ideas are, I would certainly like to see a lot more of Plan 9 brought out into the real world. There is a lot there!
      • Re:/proc on steroids (Score:3, Informative)

        by Apotsy (84148)
        If you're referring to surrogate pairs, those are a result of the insistence by Microsoft and IBM at including so many thousands of compatibility characters for their existing character sets. Everything currently in Unicode could have easily fit within 65k chars if it weren't for those two companies. The original vision of Unicode as pushed by engineers from Xerox and Apple (I know some of them) did not include surrogates.

        Also, you talk as though UTF-8 were inherently superior to UTF-16 because of the sma

    • While that's really cool and all.. I wonder if that generalization doesn't reduce the flexibility of clustering methods. I can only assume that somewhere it is configured how the machines interact. UNIX used good ole 'rsh hostname command | output' which seems to fullfill the same requirements. UNIX had many tools which were designed to be character stream oriented (cpio, tar, sed, grep) which faciliated this cross-hosting capability. But today, you have to decide whether you're using ssh and whether y
    • Right but this is not new in Plan 9 either. IBM has been doing this with the SYSOUT concept, in the mainframe world for decades. In many respects the concepts and paradigms driving Plan 9 from my adminted limity reading about and playing with the live CD seem to be a cleaver collection of everything good about UNIX coupled with everything good about SYSTEM 390/OS390/MVS/ZOS(Whatever you are calling it this year). Its a pretty cool system, but like most things despite all the industry hype it is not reall
  • by Anonymous Coward
    Plan9 runs on Xenopix DVD. Xenoppix runs Plan9 on anonymous PC.

    http://unit.aist.go.jp/itri/knoppix/xen/index-en.h tml [aist.go.jp]

    Xenoppix is a combination of Virtual Machine Monitor "Xen" and 1CD/DVD "KNOPPIX".
    It runs Plan9 and NetBSD on Xen-DomU(GuestOS) and KNOPPIX on XenDom0(HostOS).

  • Interesting reads. (Score:2, Informative)

    by ratatask (905257)
    In addition to the Plan 9 papers, here's some nice reads:
    http://www.cs.unm.edu/~fastos/05meeting/PLAN9NOTDE ADYET.pdf [unm.edu]
    http://www.collyer.net/who/geoff/9book.html [collyer.net]
  • by Lisandro (799651) on Sunday July 23, 2006 @11:44AM (#15766021)
    Great review guys!
  • "...lacks some graphical refinement..."

    Which is a polite way of saying that it's hideously ugly and designed as if the last 20 years of HCI research never happened. I mean, look at those scroll bars.

    • 1. It is a research OS.
      2. It isn't about HCI.
      3. Most of the modern GUIs also seem to care very little about usability as much as they do about marketing.
      The truth is that modern GUIs seem to be more about distraction than actually getting work done. I admit that I to love my eye candy but that isn't what Plan 9 is all about.
      Maybe that should be the next step for Plan 9. Since Plan 9 seems to be a great solution for distributed systems maybe some university should take it and start working a research projec
  • Plan9 on Qemu (Score:3, Informative)

    by int19h (156487) on Sunday July 23, 2006 @12:56PM (#15766208) Homepage Journal


    If you wish to try out Plan 9 without burning a CD and rebooting, Free OS Zoo offers an image of Plan 9 [oszoo.org] (108M) that works fine with theQemu emulator [bellard.free.fr].


    Step-by-step instructions for a Debian-based distro:

    1. sudo aptitude install wget unzip qemu
    2. wget http://www.oszoo.org/ftp/images/plan9_060327.zip
    3. unzip plan9_060327.zip
    4. qemu -net nic -net user plan9/plan9_compressed.img
    5. A window with Qemu will pop up. Press Return a few times, and you'll reach the commandline.


    Other tips:

    • Press Ctrl+Alt to toggle mouse-grab
    • Press Ctrl+Alt+f to toggle fullscreen
    • Note that Plan9 is intentionally relatively minimalistic, compared to Linux


    Good luck!

  • by ems2 (976335) on Sunday July 23, 2006 @01:27PM (#15766299) Homepage
    1) The live cd is the install cd. This isn't Linux... Installation is done by an interactive rc script ("everything is a file") in a running Plan 9 with or without rio. Try and imagine how simple it is to automate a Plan 9 installation. Unlike Linux we don't need Red Hat to develop some complex standard for doing something that should be simple.

    2) The cd comes with all the official software. Everything but the stuff that can be found in /n/sources/extra/ or /n/sources/patches/. Or anything made or ported by anyone else that can be found in /n/sources/contrib/ and elsewhere. And it definitely is not missing anything that would be basic in any operating system.

    3) It does include ping. Ping is not just limited to IP so you will find multiple ping programs for different things in their respected directories. The ping for IP is in /bin/ip/ like the rest of the IP tools (on x86 the actual location for IP's ping is /386/bin/ip/ping. /386/bin/ is bound to /bin/ during boot up on x86. Likewise /alpha/bin/ is bound to /bin/ during boot up on alpha. etc.). You use IP ping like this: ip/ping $ipadr. If you want skip the ip/ part then bind /bin/ip/ping to /bin/ping.

    4) This all fits in 80MB. Plan 9's cd is small because it doesn't have bloat. (This includes: PDF/postscript reader, page; Word processor, troff; an advance shell, rc; a web server, httpd; plus thousands of other applications.)

    5) Why didn't you ask any of your questions on 9fans before coming to your assumptions?

    6) This isn't Linux there are rules (e.g. ip tools in /bin/ip/ and http tools in /bin/http/) we don't just dump everything where ever we feel like it. What is the point of having a hierarchy without using it?
  • Plan 9 installs and runs without any problem with Plan 9 GUI under under Microsoft Virtual Server 2005 R2 running on Windows XP SP2. I have created virtual machine with 128MB of RAM. It boots directly from the latest Plan 9 ISO image. I haven't tested with Microsoft Virtual PC 2004 but I guess result would be the same. Both virtualization software can be now downloaded for free.
  • by ixra (990380)

    As an alternative to Plan9, you may try also PlanB [lsub.org], a distributed operating system based on Plan9.

  • does this still exist? When they "open sourced" it a few years ago I went to download it to give it a run through ... and it had a horrible EULA attached to downloading it. One of the most important terms (and forgive this as it's been a while so it's not verbatim) was that it was "not to be used to make weapons of mass destruction outside the US". Seemed to infer that it's ok in the US and in fact it is probably already being used thusly ...

    I clicked I Don't Accept on the web page and have never gone ba
    • hahaha... that one is long dead. The US Government required it so you should rather blame the US instead of its companies. If you look hard enough you will find that clause in a lot of places. It is still alive in the Lucent® Exptools Licensing. To quote:

      LICENSEE acknowledges and agrees that the LICENSED SOFTWARE subject to this agreement is subject to the export control laws and regulations of the United States, including but not limited to the Export Administration Regulations (EAR), and sanction re

    • That EULA no longer applies, it's legitimately open source now. That was back when Lucent still controlled it very carefully.
  • I wasn't around when it was ported, or ever learned why it was chosen, but a Plan 9 based OS called "Transit" runs on the video servers made by our company. The machines started out as general-purpose supercomputers, but after various shake-outs the hardware evolved into something optimized for storage and streaming.

    All the real fun work was done ages ago. All we see is a csh-like shell. Perl and Apache and other basic tools were ported to it, which is nice.
  • So that it can achieve critical mass. We need more competition in the OS business, especially by folks who want more than to create yet another marketer driven hack. If we had serious funding for just a few research driven OS alternatives it would make such a difference to the economy. I'd also like to see Franz at MIT get some big funding too. Maybe our next president will invest into R&D....

Uncompensated overtime? Just Say No.

Working...