On Tue, Mar 7, 2017 at 1:52 AM, Angelo Papenhoff <aap@papnet.eu> wrote:

Had they really replaced (almost?)

all code by the time of AIX 4?

I'm not saying it is not possible as it there was certainly precedent at other places. As I said, DG did a ground up rewrite for their new kernel at one point. DEC and HP switched to OSF/1 (aka CMU Mach). And I also know that the AIX base was originally considered for OSF/1 but was rejected in favor of OSF/1.

I also have not knowledge of a ground up rewrite and I did work with a number of IBM guys for a long time and my firm did a lot work for IBM. I really think if they had completely rewritten it, in the manner of DG and say Stratus; then there would have been much more notice in the community at large.

Part of why I am a little suspect is why would they have invested in a rewrite if they already had something that worked for their systems (which they did) -i.e. the business reason behind it and remember IBM was very much driven by their businesses. Stratus needed fault tolerance, so a new kernel was a requirement for them. DG, DEC were all trying to play catch up with Sun and were trying to use their new kernels as way to do so quickly.

Thus, I suspect this is an area where large sections of the AIX kernel were replaced, similar to the way BSD evolved, but can not say for sure. You would need some of the folks from Austin to chime in. I did check with some from folks from IBM and LCC at the time and validated as one of them said to me "your memory, is pretty much the same as mine."

BTW: even if they did do a whole kernel swap at between version X and Y - that would beg the questions of incompatibility. They would have had the earlier AT&T/BSD code/semantics - and would have had to specifically break it (as BSD 4.2 did in couple of cases). While possible, again, I do have memories of my brothers and sisters at LCC working on the IBM projects with a load of compatibility tests (many which we had written for them). That's not to say, there were not times when the IBM folks interpreted things differently. I do have memories explaining PDP/Minicomputer-isms to the more mainframe thinking folks. But

Clem