Announcement

Collapse
No announcement yet.

BUS1: A New Linux Kernel IPC Bus Being Made By Systemd Developers

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

  • #31
    Originally posted by uid313 View Post
    Why not just a dead simple, bare bones, minimalistic, fast text-based IPC implemented in the kernel.
    Then userspace applications can implement structured complex data passing over it using a library, example JSON or BSON.
    Because the moment you have to convert binary to text and back, it is no longer fast. A big advantage of putting the IPC in the kernel is to avoid having to copy data as much, but your approach would not only involve two unneeded copies (minimum), it would also require two conversions.
    Last edited by TheBlackCat; 10 December 2015, 07:10 AM.

    Comment


    • #32
      Originally posted by jrch2k8 View Post

      e.) nope, BSD internal are not compatible with linux at all, all pure BSD software require a linux layer to compile and in many case an ABI breaks and this is done by linux developers(mostly) not BSD ones, even cases like OpenSSH(originally for OpenBSD) require checks for other BSDs and even FreeBSD has to make their own openssh implementation like linux did

      1.) ZFS require an extensive translation layer to work on linux (SPL)
      3.) bhyve don't compile or run on linux but can run a linux Guest as any other Hypervisor, otherwise is market dead
      4.) cloudabi is focused on cloud hence linux support is a must but ok i give you this one
      5.) stated and relaunch can't even find them on Arch/gentoo repos, so I'm not sure here
      1) yes, but ZFS cannot run fully in Linux until now compared against FreeBSD and OS X, both are not the original operating system with ZFS, ZFS was ported to them.
      3) bhyve was ported to OS X and CoreOS (a linux distro), so is portable and not market dead.
      5) they are very recent projects, but by the github description both are made for linux too.

      I really didn't understand what you have talked about OpenSSH...OpenSSH as a project provides a portable version, that's cool.

      Comment


      • #33
        Originally posted by Dharc View Post

        1) yes, but ZFS cannot run fully in Linux until now compared against FreeBSD and OS X, both are not the original operating system with ZFS, ZFS was ported to them.
        3) bhyve was ported to OS X and CoreOS (a linux distro), so is portable and not market dead.
        5) they are very recent projects, but by the github description both are made for linux too.

        I really didn't understand what you have talked about OpenSSH...OpenSSH as a project provides a portable version, that's cool.
        your point was BSD guys port thing to linux and i was pointing almost neither of your examples were true in that sense, everything is portable as far as C/C++ goes, so again linux developers ported ZFS to linux, Core OS run ON top of x/bhyve(btw Mac is BSD + Mach so is easy to port between each other) but bhyve cannot run on top of a linux kernel like Xen or KVM so this is moot

        my point here is, nobody on any side of the Unix like OSes give a shit about other and they don't share code even among the same BSD communities(no netBSD developer is going to waste time making sure it runs on FreeBSD, the FreeBSD guy have to pick the code and fix it if they want it but he won't waste a millisecond checking if it work on DragonFly or Linux, etc. for example), this notion that all true Unixes are stable and ABI solid and linux is the bad guy only developing for itself is a myth.

        Also is not true all Unixes respect Posix and SysV either, Solaris kernel in the good SUN days was as different from BSD like Linux was, Every BSD spinoff out there support different hardware in different levels and in several cases drivers are not easy to port, Every BSD spinoff support a different set of features and "Posixness" and every BSD spinoff claim to be the true Unix BSD and develop only for their view of the true UNIX and in some cases hate other BSD spinoff because they aren't Unix enough like them.

        Don't confuse external Projects that take the time to make sure their code work in all platforms(like libevent or X11 for example) with cooperation between developers because there is none 99% of the time and don't confuse that certain developer from BSD or Linux helps an external project because is 100% likely he is only there to make sure his OS works first and foremost not the others OS

        Comment


        • #34
          Originally posted by Ancurio View Post

          When has Linux ever been a Unix kernel?
          you haven't read the "about Linux" ?

          Comment


          • #35
            Originally posted by gens View Post

            you haven't read the "about Linux" ?
            Last updated two years ago. Linux might be unix-like, but it is NOT certified Unix. That means they are free to deviate from the spec as much as they want.
            All opinions are my own not those of my employer if you know who they are.

            Comment


            • #36
              Originally posted by Ericg View Post

              Last updated two years ago. Linux might be unix-like, but it is NOT certified Unix. That means they are free to deviate from the spec as much as they want.
              Yep, that's the point I was making.

              Comment


              • #37
                Originally posted by jrch2k8 View Post

                your point was BSD guys port thing to linux and i was pointing almost neither of your examples were true in that sense, everything is portable as far as C/C++ goes, so again linux developers ported ZFS to linux, Core OS run ON top of x/bhyve(btw Mac is BSD + Mach so is easy to port between each other) but bhyve cannot run on top of a linux kernel like Xen or KVM so this is moot

                my point here is, nobody on any side of the Unix like OSes give a shit about other and they don't share code even among the same BSD communities(no netBSD developer is going to waste time making sure it runs on FreeBSD, the FreeBSD guy have to pick the code and fix it if they want it but he won't waste a millisecond checking if it work on DragonFly or Linux, etc. for example), this notion that all true Unixes are stable and ABI solid and linux is the bad guy only developing for itself is a myth.
                your point is wrong just a few more examples. bhyve is portable, I heard about bring it in Dragonfly times ago. HAMMER2 has portability in mind instead HAMMER and there was effort to bring it in OpenBSD in last GSoC. FreeBSD's make is the same of NetBSD and it is original from NetBSD, NetBSD folks do portable things, that's why bmake is used by default in FreeBSD, so they share codes. FreeBSD's pkg main developer pay attention on keeping portable pkg for Dragonfly (so they share codes and waste time with others), Darwin and Linux too. Dragonfly DPorts is built from FreeBSD Ports and FreeBSD Ports team accept Dragonfly patches to ease Dragonfly team. Illumos, System V derivative not a BSD, is replacing GRUB for FreeBSD bootloader. so you're wrong.

                they are not fully POSIX and SUS compliant, but they are enough for each OS share codes easily.

                Comment


                • #38
                  Originally posted by unixfan2001 View Post
                  And there we go again with the lies, misinformation and general FUD!

                  1. systemd is probably the most UNIXy concoction since forever. The fact that some people are too thick to understand this (while simultaneously screaming "UNIX!!" from the top of their lungs) doesn't change that one bit.

                  2. The reason TIPC isn't the established kernel IPC standard is mainly efficiency and a lack of progression towards modern desktop infrastructure (nor was it ever even created for that purpose).
                  For one thing, KDBUS can pass references of data between processes rather than copying the data wholesome. TIPC can't do that. The sending process can also copy data directly into the receiver's process queue, as part of a user space only, single-copy operation. Neither of those things is being offered by the current TIPC codebase.

                  nslay

                  Did you hear all that on slashdot?
                  Nothing of what you're saying is remotely true. In fact, in can be easily disproved by the dozens of software packages regularly ported from GNU/Linux to other OS'. While it's true that there are some differences in the way GNU/Linux works to, say, the way a BSD works, most of those differences are the very same kind you'll find when comparing AIX to Solaris or DragonFlyBSD to FreeBSD. UNIX isn't exactly a tight standard anymore.

                  "the most UNIXy" ... and it fails at one of the core mantra.
                  Do One Thing and Do It Well

                  Comment


                  • #39
                    So does X server. Pardon, X server does not do well what it should do.

                    Comment


                    • #40
                      Originally posted by Naib View Post


                      "the most UNIXy" ... and it fails at one of the core mantra.
                      Do One Thing and Do It Well
                      And anyone who actually understands what the point you're talking about means calls it The Single Responsibility Principle as opposed to UNIXy, which means you don't.

                      Comment

                      Working...
                      X