Slashdot Mirror

← Back to Stories (view on slashdot.org)

Linus Torvalds Isn't Looking 10 Years Ahead For Linux and That's OK

Posted by samzenpus on from the living-in-the-now dept.

darthcamaro writes: At the Linuxcon conference in Seattle today, Linus Torvalds responded to questions about Linux security and about the next 10 years of Linux. For security, Torvalds isn't too worried as he sees it just being about dealing with bugs. When it comes to having a roadmap he's not worried either as he just leaves that to others. "I'm a very plodding, pedestrian person and look only about six months ahead," Torvalds said. "I look at the current release and the next one, as I don't think planning 10 years ahead is sane."

7 of 108 comments (clear)

  1. Linux and Bloat by Anonymous Coward · · Score: 5, Interesting

    If you actually read TFA you see a little bit of nostalgia from Linus about how lean the kernel used to be and how modern Linux may be a little too bloated for some IoT applications. The truth is that Linux can certainly be less bloated that a full desktop Windows 10 installation, but it is nowhere near as lean as it used to be. Not much of an issue in larger hardware where even smartphones have more power than powerful desktops did 15 years ago, but there are definitely areas where the modern Linux kernel is a little too big for its own good.

    1. Re:Linux and Bloat by phantomfive · · Score: 5, Interesting

      To add to my earlier comment, note that most Linux installations are probably on Android, or in other embedded. If we used that as the metric, we may end up removing all mouse support, since most people don't use it.

      --
      "First they came for the slanderers and i said nothing."
    2. Re:Linux and Bloat by Dutch+Gun · · Score: 4, Interesting

      I remember a reading about metrics collected by Microsoft on features most commonly used in their Office products. What they found was that, apart from a handful of the most obvious features such as "Open", "Copy", "Paste", and "Save", the use of features flattened out very, very quickly, in terms of percentages. As such, it would really no sense to create a version of MS Word that only had the "most commonly used 70% of features", because that subset of features would tend to differ wildly from user to user.

      I have a suspicion that you'd find the same to be true of features in the Linux kernel. There are obvious features that everyone has to use, but among all the "optional" features, I wouldn't be surprised to find find that the usage curve tends to flatten out fairly quickly.

      I think there's probably a reason the most popular Linux distros are *not* the stripped down models, but the more fully-featured distros.

      --
      Irony: Agile development has too much intertia to be abandoned now.
  2. Seems Reasonable... by fuzzyfuzzyfungus · · Score: 3, Interesting

    As much as 'Linux' keeps being dragged into assorted rambling-think-pieces as though it were a direct analog to the OS-and-also-a-big-suite-of-hardware-software-and-'cloud'-service-offerings referred to as 'Windows' and 'Apple' or 'OSX'; 'Linux', so far as it is Linus' problem, is a kernel. It's also a kernel that has succeeded largely on the basis of being widely supported, reasonably flexible(with greater flexibility available to those willing to do additional heavy lifting themselves), and an inexpensive implementation of mostly-unix-like behavior.

    That's not a role 100% free of strategic considerations(like the current 'beating on the ARM vendors to un-fuck the current fragmented hellhole of disjointed BSPs and embrace sanity' initiative); but it is one where "ensure continued cooperation among interested users and hardware vendors, integrate promising out-of-tree developments as demand and maturity suggest" is more or less the best strategy to take. It's not as though it would even be meaningful for an OS to "Embrace a cloud services strategy", since that happens at a different level of the stack entirely; and to the degree that OS development does need, and do, blue-sky cool-new-architecture-from-the-ground-up; that isn't exactly mainline Linux's problem; and Linux probably isn't even an obvious starting point(if your bold new OS concept makes use of some sort of exotic hardware capabilities, you'll presumably be prototyping on FPGAs or the ASICs you are developing in tandem with the OS; if it is designed to work with mostly standard hardware; but do some part of being an OS differently, you can develop against a delightfully small and stable collection of 'hardware' thanks to VMs.

  3. Hans Reiser didn't plan ahead either by Anonymous Coward · · Score: 0, Interesting

    And where is ReiserFS now? If only Hans had done more planning and been less impulsive, his wife and his project might both be alive right now.

  4. Security is also about design by cant_get_a_good_nick · · Score: 5, Interesting

    "What I see is that security is bugs,"

    Pretty much all Outlook viruses were design issues, not bugs. They designed a mail system which, on a OS where files were executable by extension, attachments from unverifiable senders had their extension hidden so you didn't know it was an executable.

    This was baked in design. It wasn't an execution bug.

    There are entire classes of bugs you could get rid of by certain design choices. Address space layout randomization helps a lot. W ^ X, or if you can write to memory, you can't execute it. These are not infallible (there's lots of webpages on how to get past ASLR) but if we design these things as more secure, we will be more secure.

    1. Re:Security is also about design by Anonymous Coward · · Score: 2, Interesting

      Don't forget to separate the data stack from the execution stack.

      Many of C's buffer overflow exploits would be meaningless if local variables weren't stored on the same stack as return addresses. If you think about it, a return address is half of a jump instruction, yet it lies in memory that the application can write to, and so it is a violation of W ^ X.