Re: inconsistent "uname -s" option

[Robert Millan]

On Tue, Jul 15, 2003 at 10:32:13AM -0600, Bob Proulx wrote:
> Robert Millan wrote:
> > > +differ.  Some operating systems (e.g., FreeBSD, HP-UX) have the same
> > > +name as their underlying kernels; others (e.g., GNU/Linux, Solaris)
> > > +do not.
> > 
> >   "Some kernels (e.g., FreeBSD's, HP-UX's) report the same name as their
> >   overlying operating systems; others (e.g., Linux, SunOS) do not."
> > 
> > This way we are saying that the kernel "reports" that name, but we don't
> > claim that the name is correct. It's left at the readers' discretion to
> > decide wether the kernel really has that name or not.
> 
> Hmm...  I rather think that if the kernel reports the kernel name as
> XYZ then it has declared itself to be named XYZ.

That depends on what uname's options for that particular system are.
Traditionaly, "-s" has been for the whole OS but GNU/Linux didn't follow
that tradition when printing "Linux". (this seems to be the root of the
problem, but the train is far gone for that now)

>   uname -s
>   AIX

I don't have an AIX box at hand, but i guess AIX's uname manpage says
"-s" is for operating system. The dessign for non-GNU systems (generaly)
seems to be that the kernel declares the OS name instead of declaring itself.

My impression is that the whole issue is too ambigous to be able of claiming
uname().sysname will return one thing or the other. My suggestion is to
stay safe and try not claiming things that don't come assured.

> The uname command is a terrible command.

Agreed. :)

-- 
Robert Millan