On Jun 11, 2009, at 10:00 AM, Liam Proven wrote:
2009/6/11 Guy Sotomayor <ggs at
shiresoft.com>:
In reality, AIX is the ultimate Unix clone since
it's one of the
few
Unix-like OS' that is actually be branded as Unix(tm) by X/Open.
ISTR it uses a real licensed kernel, or at least, did Way Back
When?.
It depends upon which variation of AIX you're talking about. I've
dealt
with AIX/370, AIX PS/2, AIX RT, AIX V3.x and later. With the
exception of
AIX/370 & AIX PS/2, they all used completely different code bases.
What I
was referring to (because that's what most people consider these
days) is
AIX V3.x and later. Those versions only ran (runs) on RS/6000 and
later
Power machines.
The code base for AIX V3.x and later started out as some derivative
of the
AT&T code but the base kernel never was (except possibly some of the
subsystems). The control program itself was completely alien to
the AT&T
code base. The entire structure was based upon (at the time) the
Power's
nearly unlimited virtual address space (52 bits). The 64-bit
version has an
80 bit virtual address space. So the way the kernel was organized
was to
assume that address space was free. Need an array, allocate the
address
space for it statically and let the VM fault in physical memory as
it's
needed (which is also alien to most Unix implementations).
The AIX control program was mainly written by 2 folks from Watson
Research.
TTFN - Guy
Fascinating stuff. Thanks for that.
I always thought that, back in the '80s, it was a real shame that IBM
didn't keep the 386 version of AIX alive and at parity with the
others. I also wish that the portable Workplace OS/2 project had been
finished, or failing that, that they'd at least open-sourced the bits
that they did complete - it strikes me that this would have been a lot
more legally likely and doable than open-sourcing OS/2 itself, with
all the MS code in it. Workplace OS/2 ran on a Unix kernel and could
have been ported to Linux, I'm sure.