Video: Andrew Tanenbaum on Bugs and Minix' Reincarnation Server

Feb 16, 2010

Linux Pro Magazine met the author of numerous standard works in informatics and the most famous Linux critic at the Fosdem in Brussels.

Andrew Tanenbaum teaches at the Free University of Amsterdam and presented his own operating system, Minix 3, at the developer conference (Feb 7-8, 2010).

In the video he talks about his current research into stable operating systems, of the differences and similarities between Minix and Linux and his approach to the community.

Andrew Tanenbaum at FOSDEM

Related content

  • FOSDEM 2010: Andrew Tanenbaum Sets Reliability Before Performance

    Computer science veteran Andrew Tanenbaum presented the third version of his Minix operating system at the FOSDEM 2010 conference on February 6-7 in Brussels, Belgium.

    more »
  • Fosdem 2010

    A friendly atmosphere during a geek interpretation of square dance; a free exchange of technical ideas; praise from veterans: Fosdem, the volunteer-organized conference in the heart of Europe, has many faces.

    more »
  • Minix 3

    Minix is often viewed as the spiritual predecessor of Linux, but these two Unix cousins could never agree on the kernel design. Now a new Minix with a BSD-style free license is poised to attract a new generation of users.

    more »
  • Why Can't Computers Just Work All the Time?

    The feud between Minix inventor and operating system czar Andrew W. Tanenbaum and Linux Torvalds is legendary in the OS world. Before Linux there was Minix. Torvalds used to be a Minix user who set up his first Linux version in 1991 on Professor Tanenbaum’s operating system. Mr. Tanenbaum has now agreed to write a guest editorial for Linux Magazine. His opinion has not changed over the years: Linux (and Windows) are “unreliable.”

    more »
  • 17 Years Now: Linus Torvalds Introduces Linux

    Exactly 17 years ago, on October 5, 1991, Linus Torvalds sent an email to the comp.os.minix newsgroup.

    more »

Comments

  • stability

    The whole raison d'etre of a PC is the ability to program or change its behaviour via software, so a PC that has a completely fixed set of functionality, like a TV or VCR, would no longer be a general purpose computing machine. However, this programmability should be accomplished via user programs or applications, not by changing the operating system. The operating system should really never change. The main reason the operating system continually changes and evolves is because it is released in an incomplete state, ie before it is finished. One of the biggest problems with the software industry is that no software is ever finished as such - it moves from a lengthy period of continual change direct to obscelence. Or in other words, it is never finished, rather the current version is replaced by yet another continually evolving, bug ridden piece of crap with a new name.
    Linux is a classic example of modern software. There are no standardised api's (which many people actually think is an advantage - what are they smoking?), one reason why no applications can ever be written that will run on anything but the current version of the OS. Forget earlier or later versions. New OS - new applications.
    Of course with Linux the applications are actually part of the operating system in that they are bundled with it, as a 'distro' - the Linux developers have succeeded in making the whole mess into one continually moving target. If you are someone that values stability and reliability - forget computers, for now at least. Maybe in twenty or thirty years OS's will have settled down. Who knows? Would'nt that be nice.

  • Minix 3

    Good to see someone advancing OS design in 2010. His criticisms of Linux are valid and Linux does violate the most basic rules of computer science (as an un-managed system). Of course, the history of computer science is full of OS's like Minix that never become popular like Linux.

  • Reliability

    Creating an OS that is very reliable is one thing, but making it attractive to the world is quite another. There's millions and millions of issues you have to solve, like how much software does it run, how well, how fast, how many arches it supports, how much it costs, licensing issues, hw & isv support, how friendly is it (very important!), qualitative desktop, you name it.

    So just saying "we'll have a reliable OS" is too little too late. As an example look at Solaris, Sun developed the DTrace and ZFS but the market share of Solaris is still diminishing. So there's so much more for an OS to convince the industry to move to it than just becoming a more reliable Unix.

    Bottom line: the biggest problem with this future OS is the (narrow) worldview of its leader (I mean Tanenbaum of course).

  • Reincarnation Server

    The concept of a "reincarnation server" is an interesting, but I have to disagree with his statement that the user modules would not disrupt the system if they had to be restarted.

    The path leading up to the module being "hung" or failing might have set the state of the total system to a point where it has provided to the user (or to the rest of the system) an incorrect or incomplete answer. Just "killing" the module and re-starting it would not correct the over-all problem or over-all state of the system or the guarantee the right answer to the end user.

  • "Automobiles never have these problems"

    While I admit that "consumer" products like televisions and microwave ovens tend to be more "stable" then general purpose operating systems, they are built to a specific set of components, to do a specific job with specific inputs.

    As to to his analogies about automobiles, yes people do replace carburettors and other parts to get more efficiency (the "hackers" of the car world), and for reliability....perhaps he has never heard of Toyota.
comments powered by Disqus