NetBSD 10: Thirty years, still going strong

The portability is weird, NetBSD has 11 tier 1 platforms, OpenBSD has 14. NetBSD has a really impressive list of tier 2 platforms, but how usable are they?

NetBSD runs well on older hardware, even old laptops. It is clean and no-nonsense; you don't have lots of daemon processes running that you haven't put there yourself. FreeBSD is what I run more, to be sure, but for old SPARC machines that I have, I use NetBSD.

* it’s got that great *BSD flavour

* the community is good and nice

* I think the radical portability forces good design decisions - it’s a single codebase (w machine-dependent and machine-independent components, as appropriate) that runs on ~53 architectures. You get to benefit from that by using it, or studying it.

* I enjoy its fit-and-finish, as a user

* nice features like DTrace and ZFS

* pkgsrc is spawned from NetBSD and an excellent 3rd party pkg management system

* […]

This all stereotypucal nonsense though, about BSD being great for servers. BSDs are non-LTS, rolling systems, which severely limits their use for servers, where package stability is top priority. They are good for certain kind of servers, but not a replacement for Red Hat.

Why? Because it's the best for my needs. It also honors the "BSD tradition" and has fantastic backward compatibility.

I have been running NetBSD since before linux existed. Why switch to something less familiar, and, honestly, less friendly?

NetBSD is the emacs of OSs - always there, reliable, maintainable, fast, clean.

With a fantastic, friendly user community. Give it a go https://NetBSD.Org/

NetBSD seems to have more mature support for non-x86 architectures and also tends to run on old and arcane hardware.

It also just feels like classic BSD Unix, for better or for worse.

Personally I find NetBSD's pkgsrc package system to be very useful as well since it is portable to other OSes (including macOS and Linux) and doesn't require root access to install or use.

Because it works well on the Amiga.

Linux used to be the best option, but it got extremely slow, and recently even retired PCMCIA support, which is how most A1200 users access ethernet or wifi networks.

Whereas Openbsd dropped m68k support ages ago.

Sony also uses BSD for the PlayStation 4 and 5 operating systems.

https://en.wikipedia.org/wiki/PlayStation_4_system_software

Apple also used NetBSD in (some?) of their AirPort devices: https://jcs.org/2018/06/12/airport_ssh

NetBSD is fun, but I don't really see the point of it. It just seems to live in the shadows, continuing to do it's thing, producing code and funky innovations, then disappear for years, but never pointing any developers in the spotlight. Debian of the BSD world.

Netflix uses FreeBSD for their content servers. They're one of the major contributors to the FreeBSD project, focused particularly on network (eg: kernel-level TLS acceleration) and UFS subsystems.

It is not a distribution.

Short answer: The legal situation in the early 90s made BSD unattractive right at the time where cheap microcomputers sufficient to host Unix-likes were proliferating, which allowed Linux to reach critical mass instead.

Too much detail answer:

In the 80s a lot of the commercial UNIX-likes were all or partly BSD derived, like DEC Ultrix, SunOS especially pre-Solaris, pieces of IBM/ISC's AIX, etc. and by the early 90s there were a bunch of BSD ports established or in progress for up-and-coming less expensive (..at least compared to minicomputers) workstations like HPBSD, SunOS, and the Tahoe system that fell through as the target for mainline 4.4BSD, and even commodity Microcomputers like 386BSD and BSD/386 once Intel offered a part with a usable MMU.

At this point _everyone and their dog_ derived their networking stack from BSD, because it was the reference OSI TCP/IP design and all the networking parts were permissively licensed. Even the Windows networking stack is BSD derived. That's still a thing, the Nintendo 3DS and Switch's in-house OS has a network stack that is derived from FreeBSD (though the rest of the OS isn't).

Then USL v. BSDi happened in 1992. https://en.wikipedia.org/wiki/UNIX_System_Laboratories,_Inc..... and made it unclear if the BSD platform was legally viable from 1992-1994. The delays from that are essentially why Linux started and proliferated.

A few years later, MontaVista contributed a bunch of scalability work to Linux, and SGI and IBM who had recently bought Sequent contributed their even larger scaling and NUMA stuff to Linux as they bailed out of the Itanium Unix-brand-Unix Project Monterey (the later of which is what kicked off the SCO v. IBM lawsuit in 2003, but that was too late to kill inertia like the USL v. BSDi one did)... basically Linux got critical mass on features, vendor support, and hardware support by being in the right place at the right time and on relatively neutral ground relative to many long-standing divisions in the Unix world, and steamrolled the rest of the Unix market.

There was also a bunch of the common problem for permissively licensed stuff happening, in that the core folks got hired away by proprietary derivatives and choked the upstream. In the 80s a bunch of the core BSD folks left for Sun and built partially-incompatible partially-proprietary SunOS that was then superseded by the SVR4 based Solaris (SVR4 was _highly_ cross-pollinated with BSD, Xenix, and SunOS parts). Then a bunch of BSD folks spun BSDi to make commercial and partially proprietary releases (with squabbling about what would be proprietary), and the Jolitzes had a series of companies and... Then the Berkley CSRG that was the center of gravity for the BSD world closed up in 1995 (they had been winding down for years before that), and the post-4.4BSD community projects (FreeBSD, NetBSD, etc.) proliferated, with the usual open source squabbling keeping them not-very-unified.

As you note, the Next/Apple family has a lot of BSD code in it because the Mach folks at CMU derived most of their stack from BSD (and contributed back the BSD virtual memory system), and a lot of the people and code from that became the core of NeXTStep which became the core of OS X, and that lineage persists in nearly all of Apple's products and is occasionally re-synced with FreeBSD. You used to be able to get all the non-proprietary parts distributed as Darwin https://en.wikipedia.org/wiki/Darwin_(operating_system) , but Apple stopped doing that almost a decade ago and none of the forks have rooted.

There _are_ quite a number of other BSD derived proprietary OSes floating around that often don't go out of their way to note their relationship. The Sony Playstation 3/4/5 system software are hacked up FreeBSD derivatives. Juniper Junos that runs on a lot of fancy routers is FreeBSD derived (though recent releases have been migrating to Linux). Force10 (now part of Dell) and some Ericsson routers run NetBSD derived stacks. Etc.

The short story is that during dive very critical years BSDs had the spectre of AT&T lawsuit over them, plus x86 ports were less known or commercial. From my understanding this managed to get a critical mass for Linux on commodity hardware resulting in (real or perceived) situation where Linux had better chances of running well on what random enthusiast or cash strapped team might scrounge up from local used PC parts store compared to BSDs where existence of non-commercial port that doesn't require spending big money on used unix workstation.

Later on there was less difference and more fashion (unless someone dug deep into optimization or why one or the other) but arguably BSDs were still pickier about hardware they ran on.

And finally, I'd say around 2.6 kernel and NLPT, Linux actually was better for many in terms of performance - not to mention being stuff people were used to, with large software repositories, NAPI for high performance networking (at the same time OpenBSD would talk about how to physically swap your PCI NICs to make them share interrupts so interrupt on one would trigger driver receive queue for all of them), better known and corporate-tested commercial suppliers (RedHat, SuSE). For a time network heavy stuff and IPv6 favoured Free and NetBSD from my observations but that didn't last.

Probably non trivial part is that those who evacuated from commercial unices to PCs were mostly moving from System V flavour after Sun switched from BSD base, so Linux systems being closer than BSD in behaviour and API might have impacted software porting (pure speculation on my part here).

Meanwhile the only major commercial BSD based systems by late 1990s as people moved en masse to free unices on PCs was OSF/1, whose main incarnations in a larger number were Digital Unix aka Tru64 and NEXT STEP/OpenSTEP/MacOS X with all kinds of fun stuff like their BSDisms being older than FreeBSD forking out of BSD

https://debugpointnews.com/netbsd-uptime/?amp=1 Some still using BSDs

I suspect the licensing played a role in it, actually. With Linux being GPL, no single vendor could - in theory at least - take the code and make a proprietary fork. It became a neutral ground, so to speak, where corporate contributors could cooperate.

Plus the lawsuit from AT&T came at a critical time, and of course pure chance played into it as well - Linus has stated that he was unaware of the BSD projects when he started Linux, and that if he had known, he probably would not have bothered with writing his own kernel.

> I'm often surprised that the corps didn't pick up the BSDs instead of linux.

If you look at the usage, you'll find that they did, but because of licensing differences, they mostly forked proprietary versions, and only occasionally contributed back.

Yup. Our software added that bit. I've taken it out now.