The Computer History Museum has one of their long form interviews with Avie Tevanian, who led the creation of Mach (multiprocessor BSD) as his Ph.D project. (His faculty advisor went on to head Microsoft Research)

Next and Microsoft both sought to hire him, and he eventually went on to be in charge of the OS at Next and later Apple.

One of the topics he covered in the interview was how Next (and Apple) decided to use and contribute back to open sourced software at the lower levels of their stack.

He also covers what it was like to work for Steve Jobs.

https://www.youtube.com/watch?v=vwCdKU9uYnE

It's funny, for a long time Avie had a calculator on his NeXT screen to show Microsoft stock and the cost of coming to NeXT. But Apple stock did okay later, so it wasn't the disaster it once looked to be.

No, because with BSD license companies are free to withhold their code.

NeXT was never about UNIX, rather the foundation libraries, UI Builder and everything that was possible to do with Objective-C, including writing device drivers.

UNIX compatibility was a way to embrace the software of the blooming market of UNIX workstation startups, and fill a checkmark on the requirements list.

That's a good question, although I wouldn't characterize it as antithetical to his beliefs.

I'd wager it was far more expeditious to develop on top of a professional-grade OS than to start over from scratch. He wanted to sell high-end hardware, and it's hard to do that with a primitive operating system.

Anyway, interesting question, would also like to know how that decision was made.

I don't know the story, but 4.3BSD, which went into NeXTStep, was not open source. It wasn't until 4.4BSD that BSD was available without a paid Unix license.

NeXT tried to get away with a closed-source fork of GCC for Objective C, but eventually backed down under legal threat.

I think OPENSTEP for Mach 4.0, the version with the new UI that never shipped post-beta, may have been based on 4.4BSD. It wasn’t until Rhapsody that this was dusted off with bits pulled in from FreeBSD, NetBSD and OpenBSD (you can see this history in the Darwin libc source hierarchy).

Which was probably one of the reasons why they ended up writing LLVM.

The dropping of GCC and adopting LLVM is directly related to the pending switch to GPLv3 of GCC.

If there's one thing companies don't like, it's uncertainty - and the whole GPLv3 discussion gave them exactly that. They needed an alternative - and placed their bets on the more liberally licensed LLVM project.

LLVM was a UIUC project.

Ok, hiring the project lead early in development and steering the project towards Apple's goals, then.

Don't forget RMS's refusal to accept patches to gcc to run on the Mac.

The Mach kernel was never really open source, I think?

Mach kernel is open source for much longer time than that term itself exists. Original "GNU system" (what is known today as GNU/Hurd) was supposed to use Mach as kernel.

In late 80's there were multiple projects that attemped to build unix-like system on top of Mach (what essentially is academic project) out of these NeXT/OS X is the only one still really relevant. Another such project was OSF/1, which also took the way of running essentially all of real unix kernel space as one big Mach thread with code derived from BSD, this became DEC/Compaq/HP(E) Tru64 and all other systems build on this OSF/1 base never really worked with sufficient performance. (And for Tru64 on it's original Alpha platform Mach is slightly weird choice as the CPU essentially comes with it's own "microkernel" in microcode/firmware).

Yes I mis-remembered. The source was always open, but parts of it were encumbered by ATT Unix licensing.

It was heavily based on 4.3BSD, so it needed an AT&T 32v license at the time....

https://unix.superglobalmegacorp.com/cgi-bin/cvsweb.cgi/mach...