Well, it should probably be a separate thread, since the struggle went on for a long time.  We were ultimately able to integrate the AIX 3 systems with our network, but it involved computing the inverse FFT of the ODM on our master machine(s) and then disting that information as a script to the other machines.  Then we applied the updates to each machine locally.

It thus became much harder to verify that each machine was essentially identical to every other.  Of course, part of the problem was that with expensive licensed software it was easy to make an economic justification for the machines being configured idiosyncratically.  We conceptualized them as development workstations, so we wanted them to be as identical as possible.

A while later, after I got very frustrated with IBM's failure to support my distributed systems work, I left and went to Morgan Stanley where I ended up running (and evolving) the Unix environment used by the Fixed Income department.  In order to avoid single-vendor lockin we decided that ten or fifteen percent of our machines would be RS-6000s.

We wanted, however, to have an absolutely identical environment so that any user could sit down at any machine and have it just work.  Brian Redman ended up being the technical lead on that effort.  We coined the motto, "If it doesn't work on the IBMs, it doesn't work."  In a few cases we had third party products that only ran on Sun machines.  Fortunately the X Window System allowed us to run the binary on a Sun compute server while displaying a window on the user's machine (Sun or IBM).  Brian streamlined all of that to the point where there were no visible seams.

We had to standardize the user profiles, which was a bigger task than it seemed at first.

Best,

Marc