One of my first jobs I did for the company that I ended up working for decades was a job
for IBM FSD to put a second ethernet interface in to “secure XENIX” (a MLS system) to
allow the system to be used for downgrading classified material. This gave us an in
with IBM FSD and this led to us doing work with an IBM on a pair of microchannel i860
coprocessor cards called the Wizard and the W4. We ported AIX to both of them. The
cards ran inside of another AIX PS/2 system so the TCF was really handy in allowing apps
that only had 386 versions to run, the ability to maintain a common file system, and to
share peripherals. Indeed, I think the major reason IBM used the TCF concept is it gave
the 370 version of the thing an easy way to interoperate with user’s on the PS/2’s.
3270’s and other terminals designed for the mainframe really weren’t suited for UNIX.
The initial Wizard card had no I/O other than the host PS/2. It was really more or less
an academic experiment (the cards were also buggy). Two amusing things however came out
of that port. The AIX for the PS/2 had this thing to multiplex the normal VGA display
(outside of X) called the “High Function Terminal.” Our i860 version was less capable so
it was denoted the “Low Function Terminal.” The other was that I hacked the -mm macro
package to mimic the style of the IBM manuals so we could write “IBM-ish” documentation.
The W4 card was interesting. It had 4 i860 processors along with it’s own framebuffer.
One of my employees spent a lot of time of in Owego fixing the memory system (the whole
thing was set up with these Xilinx PGAs that were easy to update in the field).
Amusingly, the machine-specific parts of the W4 version of the AIX kernel had more in
common with the 370 version than the i386 version. I spent weeks out at the IBM Palo
Alto Science Center doing work on this project. I had managed to inadvertantly shutdown
the main AIX/370 in the cluster (such is a problem when things get too transparent).
Of course, while I had experience using VM/CMS before (both at the University of Maryalnd
and at Rutgers), I’d never really much dealt with the operations side of the 370. But I
found my way to a 3270 and typed “ipl aix” with at the command line with extreme optimism,
but that was indeed all it took.