Re: Debian GNU/Hurd/Windows?
Robert Millan <zeratul2@wanadoo.es> writes:
> Could it be possible to port the Hurd to Windows? I mean, to make
> Windows work as a microkernel that provides access to hardware
> and the Hurd providing a full UN*X kernel replacement.
I don't think that would be terribly useful. It might be, if you won
hardware support, i.e. if windows was ported to more platforms than
GNUMach. But both are more or less x86 only, and there are quite few
hardware things that have windows drivers but no linux drivers. (And
now I'm only talking about technical usefulness, I'm ignoring the
freedom issues...)
As for difficulty, I'm no windows wizard, but I suspect the windows
API doesn't give you enough control over a tasks address space.
Witness the implementation of fork() in cygwin. A hurd libc would need
to do similar things.
I also think that Windows IPC lacks port-dead notifications. At least
that's how I've been told that COM works; if one process keeps a
reference to a COM object in another process, and the first process
krasches violently, the second process won't notice. The reference
will never be deleted by the OS.
And there are no pthreads either, AFAIK.
So I think windows can't work well as a microkernel for HURD.
> While due to strong performance loss this would not be, if working,
> a reason for the Debian GNU/Windows port to stop, it'd be a great
> thing since GNU/Win could never provide a 100% UN*X compliant
> environment.
In theory, windows NT is a posix system, but I've never heard of
anybody using that, so it's probably more or less broken.
> Unfortunately, I lack the knowledge to implement this myself, but well
> here's my idea.. if someone wants to take it, i wish a lot of luck.
They'll need it ;-)
/Niels
Reply to: