Announcement

Collapse
No announcement yet.

seL4 Micro-Kernel Working Towards A General-Purpose, Multi-Server OS

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #41
    Originally posted by kvuj View Post
    I love what Linux stands for (user freedom / open source), but I would gladly accept a smaller & more secure kernel if there weren't too many compromises.
    smaller kernel comes with smaller(read: non-existent) userspace

    Comment


    • #42
      Originally posted by brouhaha View Post
      Geez, a bunch of people here seem to think that software reliability is all or nothing, and that having a reliable kernel is useless.
      having unusable kernel is useless

      Comment


      • #43
        Originally posted by brouhaha View Post

        Geez, a bunch of people here seem to think that software reliability is all or nothing, and that having a reliable kernel is useless.
        Hardly, the difference is that seL4 makes a very specific claim that having complete code coverage of testing makes them magically more secure and reliable than other microkernels, when the truth is that the piece of code they have complete coverage on is but a tiny portion of the stack, and so while there are advantages that microkernels provide in terms of security and reliability, seL4 isn't really all that special... and its benefit is meaningless outside of tightly controlled embedded environments where you can certify the entire stack.

        But even then that certification basically just means that that software does what the Requirements Docs say it does. If the requirements documents say "Send passwords in clear text" then as long as it does that then it's "proven".

        Comment


        • #44
          Originally posted by Luke_Wolf View Post

          Hardly, the difference is that seL4 makes a very specific claim that having complete code coverage of testing makes them magically more secure and reliable than other microkernels, when the truth is that the piece of code they have complete coverage on is but a tiny portion of the stack, and so while there are advantages that microkernels provide in terms of security and reliability, seL4 isn't really all that special... and its benefit is meaningless outside of tightly controlled embedded environments where you can certify the entire stack.

          But even then that certification basically just means that that software does what the Requirements Docs say it does. If the requirements documents say "Send passwords in clear text" then as long as it does that then it's "proven".
          Great, suggesting other microkernels? Which ones? Are better? Go on ?

          Comment


          • #45
            Originally posted by Luke_Wolf View Post



            Congratulations on being scooped up by marketing. No It's literally just as I said. Oh boy you have 100% code coverage on your tests which means you were following TDD. "Mathematically Proven" is as much a bullshit term as "Artificial Intelligence". It is only suitable for annoying actual developers.
            If you don't know what a formal proof is, you don't know, but don't go telling others that they got it wrong when you literally have no idea what you're talking about. The concept of "test coverage" doesn't even apply to this, and they aren't even tests. A formal proof is a real tangible thing, unlike a marketing claim of "artificial intelligence"

            Luke, go do your homework, you're making an ass of yourself.

            Comment


            • #46
              Originally posted by pal666 View Post
              having unusable kernel is useless
              Indeed, 30 millions lines of code. Hard to maintain.

              Comment


              • #47
                Originally posted by pal666 View Post
                smaller kernel comes with smaller(read: non-existent) userspace
                G E N O D E. Xen. Qubes. QNX. Android. Linux. All of this. is my userspace under a microkernel. How do you eat yours?

                Comment


                • #48
                  Originally posted by brouhaha View Post

                  Geez, a bunch of people here seem to think that software reliability is all or nothing, and that having a reliable kernel is useless.

                  Comment


                  • #49
                    Originally posted by Luke_Wolf View Post

                    Ah yes because it's not as if a Network Protocol Stack contains information that would potentially be bad to have leaked or anything, which is what Bleeding Tooth is as much about.
                    No it isn't.

                    Comment


                    • #50
                      Originally posted by bofh80 View Post

                      Great, suggesting other microkernels? Which ones? Are better? Go on ?
                      Minix, QNX, Singularity, (in the future) Redox... The list can go for quite some time. The L4 kernels are hardly the only entry in the market and they're far from the most interesting or "complete". They were useful to prove that it was just a Mach fuckup that microkernels were considered slow, but at this point their main argument in favour of themselves is mastubatory marketing crap that doesn't actually mean much of anything.

                      Comment

                      Working...
                      X