Who was Holly Knight in the passwd files?

holly:lLoTSq8z2Qmuw:8:8:Holly Knight:/usr/holly:/bin/csh
holly:.DhxL2fe01f5s:8:10:Holly Knight:/usr/holly:/bin/csh
holly:GSNDZRS19RO/g:8:1:Holly Knight:/usr/holly:/bin/csh

Also – 30 years later my fingers still sometimes type “sync” when I’m idle and thinking 🙂

Yeah, you can see the kernel event manager I’d written there – it sort of mirrors the mac one, the Apple people took that and integrated it into their toolbox (I’m not sure how much of that survived into 3.0) – essentially it’s a queue of events (mostly mouse and keyboard), it also does software mouse movement (there’s tight assembly code to hide the mouse, move it, unhide it)

Author of Shoebill here — really cool to read about your work porting Sys V to the mac. I had a lot of fun reverse engineering your code 😉

E.g. https://imgur.com/a/GDc4e1S

Yeah, prompts were a big deal – they still are (I’m working on a highly OO/superscalar/speculative risc-V project, getting a prompt a couple of weeks ago on an AWS FPGA instance was a big deal!)

We used to quote 6 weeks for the above process for a system with a known MMU and the two drivers – fastest I ever did was 3 days for a system where we’d already done all the bits before for other customers – marketing made me wait a couple of weeks before telling the customer.

A/UX had a lot lot more work than a bog standard port, and it took a lot longer

sounds pretty cool!

Best I could do was polled MPU401… it’s a lot harder when you dont know what you are doing and are guessing at how things work.

It must have been really rewarding to see that # prompt!

We used an 11/750 to do the initial bootstrap process when porting to new platforms, but always switched to building on the box as soon as it became stable, building that way was always much faster than downloading kernels over serial lines for testing.

Porting sort of went this way:
– write a polled serial driver, and a polled disk driver
– cross build all the apps
– integrate the drivers into a standalone app that could download a boot file system
– set it downloading, go away for many hours
– write interrupt driven disk/serial drivers, build a kernel
– start downloading and debugging the kernel
– get a prompt (big milestone!)
– as soon as it’s stable move to development on the box

Did you cross compile from something like a VAX? I can’t imagine using the target to do builds back then it’d be so incredibly slow