CIRCLE WITH A DOT

@merlin The UnixWare name came from Novell, before SCO acquired it. Novell bought it from Unix System Laboratories.

@Natasha_Jay I suspect that one of the few applications where a heavily hallucinating LLM can outperform a human would be in replacing board members, C-suite executives, and their direct reports. I propose a longitudinal study with a control group of high level executives using real data, and experimental group using hallucinated, or maybe even totally random data filtered into plausible ranges, using executive compensation deltas as the metric.

Wondering if plain-old wire-wrap connections can even handle the kinds of clock rates that would justify using 74F series logic instead of plain-old low-power Schottky. Would I have ended-up with a nightmare of crosstalk and RFI, both incoming and outgoing?

In theory, 74F logic should handle clock-rates up to 100 MHz, but I'm guessing it would take years and years of fiddling to get it stable at even 1/3 of that. Fairly sure PCB prototypes that big would cost in the triple digits for a small run, for each iteration, and several iterations would almost certainly be needed.

And in the end, it's a very complicated toy that's less useful than the simplest microcontroller, until a similar amount of effort goes into all the other boards required before it will actually work as a computing device. At which point, it will still be an inert box. It would need a cross-assembler hosted on something else, then it would need a ROM monitor, before I could even think about trying to port something like the XV6 kernel to it, then cobbling-together a bit of a userland and a native assembler. A self-hosted compiler is probably a bridge or three too far.

Of course, if I'd thought to try and find a few other people with the same weird obsession and we each took on a different part of the design, it might have been possible.

Ahh, the arrogance of (comparative) youth.

Long, long ago, probably after reading _Soul of a New Machine_, I had a naive notion of stringing a bunch of 74F181 and 74F182 bit slice ALUs and carry units together, opposite several million words of SRAM on the biggest protoboard I could find, then incrementally bodging-on a bunch of other chips until I had a working CPU board for a new architecture of minicomputer. Mini, as in fits in just a couple of 4 post cabinets, rather than occupying an entire room.

Not sure why I thought I would be able to focus on a single project for long enough to actually do that, despite all precedents to the contrary. The accumulated dust on the tubes speaks volumes. Also not sure if I ever thought ahead to interrupt handling, DMA, additional RAM, storage host bus adapters, networking, floating point, or any of that other stuff which might permit the CPU board to actually do something. There were two tubes of old UARTs rubber-banded to the ALUs and carry chips, though, so clearly, I was planning on a whole bunch of serial ports.