On Mon, Mar 6, 2017 at 11:20 AM, Joerg Schilling <schily(a)schily.net> wrote:
The kernel definitely looks like it was written from scratch.
The fact that there are functions like "issig()" make it obvious that the
authors did have access to at least the V6 kernel via the Lions book.
I can say for a fact that is not how it was. AIX was a port starting
with AT&T code originally targeted to run on DEC and Intel Systems.
As I said previously when I ran down the history of AIX, the developers had
AT&T licenses. As I was reminded in an off line discussion with one of
the IBM guys when I was checking to make sure, ISC did the original 386
port for all of AT&T, Intel and IBM (one port - 3 checks). ISC also
started the AIX port, with a number of the folks moving to LCC which was a
step I left out in my previous email sorry, since it was implied when I
said they started with that AT&T 386 stuff (which AT&T got from ISC).
Bottom line.... it was not a rewrite, it was always a port.
One of the issues that I think many people from the outside looking in, do
not fully understand is that many firms struggled with was how difficult it
was/is to keep things current (BTW: Linux struggles with today just as
much as UNIX ever did - just less marketing $s being spend - it is very
hard problem).
Companies like DEC, HP and IBM start working with one version of the kernel
or worse yet, the command system and enhance it as they need. But time
moves forward and their version and the rest of the world start to become
different (branch/fork). Linux has been mostly able to keep the kernel
the same, but not the command system.
Anyway, the question is how do keep your "branch" current. Not only do
you have to compete with other "vendors" -- you also have we now call the
FOSS community creating and enhancing the tools, so you want to pick up
those new tools and or some if not all of the enhancements to those you
already have. Plus those enhancement are likely to conflict with you own.
Its a struggle and the bigger the firm, it seems like the harder time
they have doing it.
General you start with one, and just keep folding in. Rarely do you swap
out. Both DEC and HP used OSF/1 as time to swap out the command system.
DEC swapped out the entire kernel, HP did not. Interestingly enough, IBM,
who's license OSF was using, IIRC did not use either part from OSF in it;s
entirety, but instead took things back piece by piece.
Anyway, I think AIX as a whole was an example of that. I'd have to check
with some of my old LCC coworkers about what versions of the command system
was used to start with for AIX. I never directly worked on that project so
I'm personally not sure. My guess it would have been PWB III time frame
was the SCCS starts, with a lot of BSD injected because of the University
focus, whatever was kicking around Yorktown, plus whatever ISC had, plus
whatever LCC/UCLA had - oh yeah and originally it had to run both on 386s
and 360s, so the code user space had some stuff in that was "different"
from what you saw on Vaxen.
IBM Marketing (just like DEC and HP marketing) technically decided what was
in or out, not the techies (although at DEC we were probably a little more
devious). But at DEC not matter what was in the code - the "SPD" was the
final statement - and that was own by Marketing. I'm pretty sure AIX
worked similarly.