Dear unlucky friends,
Frank Heckenbach writes:
You will never see a GNU/proprietaryKernel - this is a contradiction to the goals of the GNU project.
So if I use a Linux kernel with GNU tools I should give credit to GNU by using the term GNU/Linux. But if I use a Solaris kernel with GNU tools, I *must not* do so?
Perhaps it would help to distinguish between the 'Operating Environment' and the Operating System. OE would be defined as all Operations that can be invoked via some API after establishing the environemnt. The GNU shell environment would be i.e. established upon login into a linux box (or after calling 'set-gnu-environment' on an IRIX box) and have GNU make as make, gnu sed as sed and so on.
The concept 'OS' is less well defined, since it has userspace parts like administrative programs or the shell.
Giving credits to someone should not be the issue here (as some seem to think), but functionality (does 'make' do what 'gnu make' does). Should I call a Linux system GNU Linux then? I see 3 threads of argument here:
1) Linux incorporates many GNU Tools. -- I think, this doesn't apply, since it also incorporates X11, BSD-Tools and so on. I'd have to call it Qt/GNU/X11/BSD/Linux.
2) The system provides a GNU-ish Operating Environment, that is, on , whose structure follows a philosphy -- I can't comment on that, since I haven't read enough GNU papers. Perhaps someon could point me on a paper, which states, what the GNU system structure will be? The GNU coding standards are certainly violated often enough in Linux, so it shouldn't be ALLOWED to be branded GNU :-).
3) Credits. Since someones tool T, helped someone else to make X, someone else should call X T/X, to give credits. -- Hm, yes I think, credits should be given, no problem, but not in the name. I once used a Borland compiler to write a boot loader. Should I have called it Borland/MyLoader? Hardly: Trademark rights would have got me soon enough.
Im really puzzled, but please don't flame now. The only valid claim IMHO would come from 2), which also would imply some minimal Gnu libc/shell/whatever API standard to which a Gnu system must conform. Can anyone point me to something like that (at least a rough draft).
I really think Gnu should/must be some kind of branding like POSIX or Unix98 and so on, if it should have any value at all. Having Redhat print GNU/Linux at their boxes (or SuSe) and then shipping anything they like (as they do: look at the patched kernels, differnt admin tools, different build systems, non portable RPMs) will not help me.
Don't sell the 'Gnu' word too cheap!
Regards -- Markus
PS: I think this debate is really dead and over. Please don't answer, if you *only* have an opinion, instead of an argument about (a) the validity of the arguments 1-3 (my logic might be deficient here) or (b) new empirical data (like a written Gnu definition). I'm only contributing this mail since I'm really intrested, wether there is actually a definition for a Gnu OS or Gnu OE.