> we might be providing an environment to run other operating systems
> binary's, but we are *not* that other operating system.  i'm yet to
> see any argument except "breaking <random software>" (which i do
> consider a valid argument, but not an encompassing one).  i like the
> idea of providing the ability to change the utsname returned, per-host.

Breaking random software is exactly the point.   As several people
have pointed out, we really need a /emul/*/bin/uname that reports the
correct operating system so that O.S.-specific install scripts work.
RPM should work - I should be able to install Linux apps without
losing horribly.   Right now, when I try to install a Linux app, it
says ``unknown operating system.''   Don't tell me to recompile RPM
and hack the install scripts - that's the wrong answer for Joe Random
User, even if it could work for me.

> also, *what* version do you decide to return?  what if i have some
> <random software> that expects "FooOS 3.4" but our emulation for FooOS
> returns 3.5?

We should probably be trying to emulate a particular version of the
operating system, in which case we'd have a version number to report,
even if not the one a particular app would like to see.

It's unlikely that an application other than an LKM is going to do
what you describe above - on Linux, at least, you pretty much have to
test for features because there are so many random releases out there.
If you try to match on a particular version number, you are pretty
much guaranteed to lose.

			       _MelloN_