On Thu, 2004-04-01 at 08:25, William Donzelli wrote:
> > Umm, no. AIX has nothing in common with OS/400.
>
> My source is "Introduction to the RS/6000", and official IBM document from
> the very beginning of the line. If you give me some time, I could find the
> book and give the IBM number. It is a thick engineering/sales freebie, and
> has quite a few details of the internals of POWER and AIX (although not
> enough to do anything fun). It states that a some of the aspects of AIX
> were taken from OS/400 (I think aspects of the file system, but don't
> quote me on that until I find the book).
Maybe some design concepts, but there was *no* code in common.
>
> > It is true that the AIX kernel (on POWER & PPC) was a custom written
> > control program (written at T.J. Watson Research Center) with UNIX
> > semanitics layered on top. This was to take full advantage of the
> > POWER's architecture (especially in the VM area) that would have been
> > too much work to adapt a "standard" unix kernel to.
>
> Well, sort of. AIX has a custom kernel, but was designed to be easily
> ported to other architectures - namely Intel architectures. This is also
> mentioned in the above document. This actually happened with the T386 and
> T960 router cards, used with the old NSFnet RS/6000-T3Bs. Each router card
> runs a cut down AIX on 80386 or 80960 microprocessors, and hadles all of
> the routing duties - the RS/6000 is just there for the ride,
> basically. These routing cards today are rarer than hen's teeth (look for
> extra thick MCA cards (T960), or even extra tall ones (T386)).
No, the AIX kernel was *not* easily ported to other architectures.
That's why OSF chose Mach for the OSF/1 kernel. I've been through the
AIX kernel source (used to work in Austin) and I can tell you it is
*very* specific to the POWER architecture. The VM (which permeates the
entire kernel -- address space is free) is welded to the POWER's MMU
design.
--
TTFN - Guy
Received on Thu Apr 01 2004 - 16:23:27 BST