PDP-8 Unix was Re: Who wants a nice monitor (UK)?

On Tue, Feb 7, 2012 at 12:51 AM, Cameron Kaiser <spectre at floodgap.com> wrote: The oldest/weakest I've personally run UNIX on was an 11/24 w/dual RL02 (10MB each) and 2MB of RAM. It's far from a record. If I can find an affordable EAE for my PDP-11/20 and figure out how to sub the swap disk with something other than an unobtanium fixed-head disk, I'd run it there, but that's a long-term project. UNIX? Really? Or just something with a Bourne-like command line? I gotta say that I don't think so either. I am not aware of a C compiler for the PDP-8, but I think it would be exceedingly difficult to write one if were possible at all (I've done a lot of system and embedded code in C and debugged it in assembler, so I know a bit about the syntactical mapping that goes on with C abstractions to the instruction sets of MC68000s, VAXen, PDP-11s, etc). I would like that too. I've even thought a lot about it. IMHO, anyone who does it will be writing the Z-machine in PDP-8 assembler, just like what was done for the Z-80 and the 6502 for 1970s and 1980s micros. I _think_ the Mac Z-machine was the first written in C, but I could easily be mistaken on that. I know there were official Z-machines written in C for the Mac and the Amiga, and probably the later ones on the PC (not sure about the early v3 interpreters for DOS). I like Frotz. Frotz is huge compared to the 6K-8K early 8-bit interpreters. It gets more dicey trying to ask a 12-bit machine with 4K pages to emulate a 16-bit virtual machine. I would consider it a win if one could fit the Z-machine code in 2 fields with enough space left over for a 2-page system handler and a 1-page line printer (SCRIPT) handler, using any memory above 8K for object data and game file buffers. Three fields seems plausible. A few years back, I assisted with a modern from-scratch Z-machine effort for ElfOS on the 1802 (that I was showing off at an early VCFmw). On a 32K Spare Time Gizmos Elf2000, once the interpreter was loaded and the object tables were loaded, there was very little room to buffer the game - I think it was on the order of 1-3 512 byte disk blocks. You can fit a v3 game and interpreter in 32K, but to do it in less would probably require a read-write virtual memory scheme on the object data (fortunately, a full boat on a PDP-8 is 32K 12-bit words not 32K 8-bit bytes - that helps too). Strangely enough, I was just thinking about a 12-bit Z-machine this week. Anyone out there have 12-bit coding experience and have time to answer a few questions about OS/8 and file interchange from the 8-bit outer world? -ethan