that old GNU/Linux argument

[Les Mikesell]

Andrew Kelly wrote:
> 
> <snip for brevity>
> 
>> Except that what you call Linux software isn't actually for Linux.
>> Have you ever heard of Nexenta (GNU/kOpenSolaris), Debian
>> GNU/kFreeBSD, and even UnixWare?
>>
>> If you take *GNU* libc, rebuild it to target a different kernel while
>> exporting the same ABI, and voila, you can drop Linux entirely from
>> what you call a Linux Operating System, and pretty much all
>> applications will still work just the same.  Because they're not
>> applications for Linux.  They're applications for GNU libc.  They
>> couldn't care less that they're running on top of the kernel Linux.
>> For them, the kernel is irrelevant.
>>
>> Next frequently raised fallacious objections?
> 
> Ladies and gentlemen, I give you "the sound of a shoe dropping".
> [applause]
> 
> That was a fabulous exchange, very well stated, very well explained.
> 
> Alexandre, very impressed with your ability to avoid vitriolic and
> abusive tone. Very impressed.

Except that he didn't go quite far enough with that explanation.  Not 
only is Linux just one implementation of the more or less standard 
Unix/Posix system call interface that predates it, but so is GNU libc 
just another implementation of the pre-existing standard c library 
specification and sensibly written programs have no dependencies on any 
specific implementations of these standards.  From his description you 
might think that it would make sense to say GNU/apache or GNU/sendmail 
when in fact, like most such programs  there is no such relationship and 
they run just fine when built on *bsd or commercial unix C libraries - 
as they were before glib existed.

It would make more sense to describe a lot of things as the "GNU 
re-implementation of..." rather than to imply that they were created as 
original designs or are the only versions that exist.

-- 
   Les Mikesell
    lesmikesell at gmail.com