inconsistent "uname -s" option

[Robert Millan]

Hi!

I've found that the -s option in GNU uname has some serious inconsistency
problems. From --help output:

  "-s, --kernel-name        print the kernel name"

The first problem is:

  With a single exception, this isn't true for any system: "HP-UX", "AIX",
  "FreeBSD", "NetBSD", "OpenBSD", etc.. they're all complete OSes and not
  just kernels. But GNU uname running on any of these will say they are
  kernels.

  The exception is GNU/Linux, since it prints "Linux" which is the kernel. This
  is due to the well-known fact that Linux' developers refused to give credit
  to GNU when they started building a Linux-based GNU system. The correct
  solution would be to fix uname () to return "GNU/Linux" as sysname, in
  either Glibc or the kernel Linux. I don't think this is technicaly viable,
  though.
  
The second problem is not adhering to standards:

  - POSIX says "-s" should:
    "Write the name of the implementation of the operating system."
  - And you're using uname () system call to obtain the string, which
    according to POSIX is expected to provide:
    "char sysname[]  Name of this implementation of the operating system."
  - And particularly, the GNU C library's specification on uname () says:
    "char sysname[]  This is the name of the operating system in use."

I propose two solutions:

  - Make the --help output dependant on the platform, and only say "-s"
    is for the kernel when running uname on GNU/Linux. This has the advantage
    that we'd be telling correct output for all systems but at the inconvenient
    of having to maintain documentation for different options on different
    platforms. I can send a patch for this if you like.
  - Revert it to the old behaviour. This has the only disadvantage that uname
    would say "Linux" is the system. But GNU uname is just a messenger; if
    you really want to fix this you should discuss it with the maintainers
    of Glibc and/or the maintainers of Linux.

Thanks for your time,

-- 
Robert Millan