This OS was implemented on DG minis in about 1970 by Educational Data Systems.
Dan Paymar, one of the EDS cofounders, hired me to help him complete a one-user diskless Basic Interpreter in Summer 1969. I remember debugging the MAT INV operation by listening to the funny patterns of noises the FM radio made as MAT INV ran; you could hear it processing matrix rows. We did this on Data General Nova serial #3, using a truly awful teletype-based text editor and assembler that read and wrote paper tapes. We produced mountains of "chad" from paper tape holes, ugh. And you haven't lived until you've made a 1000' paper tape and the 3 year old kid belonging to your boss tears it in half.
EDS wanted to sell "educational" computers to schools but one terminal wouldn't serve multiple students so we built a 4 user memory-only standalone Basic interpreter with bad but workable timeslicing in 16Kb core memory. Hardware multiplexors for multiple bit-serial devices (in our case, Teletypes) barely existed and were really expensive, so we built a dirt-cheap 16 bit parallel output register with one bit per terminal, and I built a device driver using the 1000hz built-in interrupt to serialize the ASCII bits being sent to the output ports. This continuously burned about 35% of the CPU even when nothing was being transmitted, but still left plenty of CPU for the timesharing Basic. We could drive 16 Teletypes full blast (110 baud) at the same time, making one hell of a racket.
In Summer 70, the company got a head-per-track disk drive, and (zounds!) 32kb of core memory, and Paymar asked me to write a version of the BASIC that stored program files on the disk. I was a bit contrarian and being left unsupervised :-} I inverted the problem to produce a multi-user disk operating system that would would run a BASIC interpreter as one of many applications, initially called Alice. The head per track drive meant Alice could swap users remarkably fast and the resulting response times to the teletypes was still quite good. The disk also gave us space to build a respectable file system where files actually had built-in record description metadata. Alice could also run a text editor and an on-line assembler using disk files so we could finally avoid using the paper tape reader and high speed punch! All of this is done in hard core DG assembly language. My college buddies Steve Sternitzke and James Neighbors were a major part of this exercise.
This was frankly pretty impressive: a real timesharing system on a minicomputer. Yep, earlier than Unix. I've been stunned for my entire career that I got do to this when nobody was looking.
Six months later, EDS ran into a financial troubles and I went on to do other things (eventually building an incremental BASIC compiler for Keronix DG clone computers and other operating systems https://retrocomputing.stackexchange.com/a/4541/6744). EDS renamed the OS to IRIS and started to sell it as a Business Basic system. Apparently it was pretty successful in that market. I came back as consultant a number of times, the last time sometime in the early 80s to help them figure out how to get 64 users running on that a DG clone called "point 4" (after the .4 ns cycle time) now made by EDS renamed to "point 4 corp". It was rather stupid to name the company after the cycle time of a product... sure enough, it got faster but the company name was stuck. Marketing people... sheesh.
I don't have a good timeline of what happened after I left. Any more stories of how it evolved would be interesting.
