Re: [RFC] hurd/hurd_types.defs

[Erik Verbruggen]

On Tue, May 01, 2001 at 01:52:43PM -0700, Thomas Bushnell, BSG wrote:
> Neal H Walfield <neal@cs.uml.edu> writes:
> 
> > I find it incredibly disturbing that hurd/hurd_types.defs defines off_t
> > and size_t to a type int.  What bothers me even more is that MiG does
> > not define a type long.  What is the correct approach to fix this?
> 
> Lol.  We can't. :)

So much for Mach/MIG. ;-)

> A MiG int is 32 bits, always and everywhere, on every single
> implementation, no matter what the native format is.  This is
> absolutely necessary to achieve network transparency.
> 
> At some point we may well want a longer off_t.  (Though not
> necessarily; we could just as easily change the interfaces to take two
> ints instead of one.)  But remember here that MiG definitions are
> establish network interfaces, with fixed bit widths, and not merely
> local machine things.

Um, well, we might want to update this. Or something. Or just drop the
network part. Just out of curiosity: Mach did work on Alpha, no? How did
they do this 32/64 bit stuff there?

Erik.