NetBSD 10: Thirty years, still going strong(fosdem.org)
fosdem.org
NetBSD 10: Thirty years, still going strong
https://fosdem.org/2024/schedule/event/fosdem-2024-3508-netbsd-10-thirty-years-still-going-strong-/
5 comments
I would like to know from anyone that runs NetBSD, a simple question: why?
Any and all answers appreciated, but for a very long time, I've boxed NetBSD into a corner of "extreme portability" without much else to hold it as practical. FreeBSD is great for all around general purpose desktops and servers, OpenBSD is great for the network edge, and NetBSD... I guess you can run on your Dreamcast.
Any and all answers appreciated, but for a very long time, I've boxed NetBSD into a corner of "extreme portability" without much else to hold it as practical. FreeBSD is great for all around general purpose desktops and servers, OpenBSD is great for the network edge, and NetBSD... I guess you can run on your Dreamcast.
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/
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/
> I have been running NetBSD since before linux existed
Linux predates NetBSD
Linux predates NetBSD
I assume PP either is a 386BSD developer or meant *BSD. ;-)
* 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
* […]
* 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
* […]
I hadn't realized DTrace and ZFS were ported to NetBSD. I think those two alone imply someone is using it for heavy-duty systems, nice. :)
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.
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.
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.
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.
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?
Admittedly, I hadn't looked recently, but that values does seem flipped around from how I remember it (then again, tier 1 was probably always a lot more minimal than tier 2).
For comparison, FreeBSD has just two tier 1 platforms: https://www.freebsd.org/platforms/
OpenBSD has 14, as you mentioned: https://www.openbsd.org/plat.html
NetBSD, I count 9: https://wiki.netbsd.org/ports/
FreeBSD has an additional 14 platforms in tier 2 or 3 support (counting FreeBSD 13; 8 platforms counting only FreeBSD 14), OpenBSD has no non-tier1 platforms, and NetBSD has too many to count, but it's notable that most of them are updated to the current release.
For comparison, FreeBSD has just two tier 1 platforms: https://www.freebsd.org/platforms/
OpenBSD has 14, as you mentioned: https://www.openbsd.org/plat.html
NetBSD, I count 9: https://wiki.netbsd.org/ports/
FreeBSD has an additional 14 platforms in tier 2 or 3 support (counting FreeBSD 13; 8 platforms counting only FreeBSD 14), OpenBSD has no non-tier1 platforms, and NetBSD has too many to count, but it's notable that most of them are updated to the current release.
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.
>which severely limits their use for servers
Ahh i found the Oracle user. Redhat has not "more stable packets" but "more known bug's and eventual workarounds and less packets".
>They are good for certain kind of servers
Like Redhat is just good for a certain kind of servers, like that OracleDB for your Delphi-App ;)
BTW: I think 5 years for the base-system is LTS (FreeBSD) and NetBSD even longer, and no BSD's are not rolling-systems hence the release 10. The packets however are rolling or quarterly (for NetBSD and FreeBSD), and ~point for OpenBSD.
Ahh i found the Oracle user. Redhat has not "more stable packets" but "more known bug's and eventual workarounds and less packets".
>They are good for certain kind of servers
Like Redhat is just good for a certain kind of servers, like that OracleDB for your Delphi-App ;)
BTW: I think 5 years for the base-system is LTS (FreeBSD) and NetBSD even longer, and no BSD's are not rolling-systems hence the release 10. The packets however are rolling or quarterly (for NetBSD and FreeBSD), and ~point for OpenBSD.
This is pointless argument, as you do not understand why industries like stale-like-rock packages of Debian and Red Hat LTS systems. No one cares much that base is LTS, what is important that third party _packages_ stay same for years and only receive security update. Essentially zero-masintenance install and forget for two years systems, not quarterly update-reboot-fix-breakage every 3 month. No go fo many server usage scenarios.
True 25 years in the industry and i still have not learn'd that actively maintaining and continually upgrading and testing systems is the better path then to install rocks in the closet and forget about them...then have to upgrade from PostgreSQL 7.5 to 15 in one week, oh and can you have a look at our nfs v1 server?
Damn our software is not working with Rhel9, can you ask that Software producer if they have it for Rhel9? What? They are out of business since 3 years? Ok let's ask Redhat if we can extend the support for another 5 years, and let the software rebuild..i calculate 10 millions, what about you?
Look Redhat is really good for IoT light-switches and such, with a expected life of 10 years..but not much more...oh and bank and insurance...but that's another story.
BTW: Do you think "Security-updates" are no-risk updates?
Damn our software is not working with Rhel9, can you ask that Software producer if they have it for Rhel9? What? They are out of business since 3 years? Ok let's ask Redhat if we can extend the support for another 5 years, and let the software rebuild..i calculate 10 millions, what about you?
Look Redhat is really good for IoT light-switches and such, with a expected life of 10 years..but not much more...oh and bank and insurance...but that's another story.
BTW: Do you think "Security-updates" are no-risk updates?
I worked in IT for 20+ years too, so you do need to flaunt your experience around.
Yeah, well, this is why you have 2 year update cycle, like Ubuntu, Debian or you have overlapping support periods like Redhat. You can update you system every once in a while, you know beforehand what will be in the next release and when it will happen. Not some insanity of rolling or semi-rolling packages like BSDs have.
Yes, security updates are always a lot less risky than quarterly FreeBSD updates. Quarterly does not receive all sec updates anyway, but do receive some. Worst of both worlds.
In fact I think, FreeBSD is not suitable for majority server purposes exactly due to terrible support cycle, despite clear technological advantages (ZFS, Jails etc.). If you limit yourself to some narrow uses, completely serviced by the base system, then yeah you may use it as server, but that is very niche use.
FreeBSD actually is a good replacement for Arch-like systems, for workstation uses, exactly due to its semi-rolling nature and separate base systems, this will cause less breakage compared to Arch.
Yeah, well, this is why you have 2 year update cycle, like Ubuntu, Debian or you have overlapping support periods like Redhat. You can update you system every once in a while, you know beforehand what will be in the next release and when it will happen. Not some insanity of rolling or semi-rolling packages like BSDs have.
Yes, security updates are always a lot less risky than quarterly FreeBSD updates. Quarterly does not receive all sec updates anyway, but do receive some. Worst of both worlds.
In fact I think, FreeBSD is not suitable for majority server purposes exactly due to terrible support cycle, despite clear technological advantages (ZFS, Jails etc.). If you limit yourself to some narrow uses, completely serviced by the base system, then yeah you may use it as server, but that is very niche use.
FreeBSD actually is a good replacement for Arch-like systems, for workstation uses, exactly due to its semi-rolling nature and separate base systems, this will cause less breakage compared to Arch.
You still don't understand what i meant. Breakage means you have not tested anything, it's your fault and your responsibility.
To stay near upstream is incredible healthy, software stacks are a living thing and should be treated like that, remember the Mainframe trap?
FreeBSD supports for example Java8, PostgreSQL12 and GCC4.8.
The port-tree has not just the latest upstream version available.
And why would i want a Red-hat-patched 6 y/o Nginx on my machine? No thanks.
>Yes, security updates are always a lot less risky than quarterly FreeBSD updates
Let's stop here when you start comparing apple with oranges.
To stay near upstream is incredible healthy, software stacks are a living thing and should be treated like that, remember the Mainframe trap?
FreeBSD supports for example Java8, PostgreSQL12 and GCC4.8.
The port-tree has not just the latest upstream version available.
And why would i want a Red-hat-patched 6 y/o Nginx on my machine? No thanks.
>Yes, security updates are always a lot less risky than quarterly FreeBSD updates
Let's stop here when you start comparing apple with oranges.
If - for some reason - FreeBSD would not be available - I would be running NetBSD.
The things I miss in NetBSD (having them on FreeBSD) are:
- ZFS Boot Environments with beadm(8) command.
- Bhyve (I hope NVMM will get to similar level)
- Jails - not sure NetBSD will bring any containers into the 'game' ...
- Probably some other things that I missed because of not not running NetBSD daily.
I heard WINE is not a 1st class NetBSD citizen - so that helps a lot here.
IMHO (with my limited NetBSD experience) seems to look bright ... and I wish NetBSD and NetBSD community all the best.
The things I miss in NetBSD (having them on FreeBSD) are:
- ZFS Boot Environments with beadm(8) command.
- Bhyve (I hope NVMM will get to similar level)
- Jails - not sure NetBSD will bring any containers into the 'game' ...
- Probably some other things that I missed because of not not running NetBSD daily.
I heard WINE is not a 1st class NetBSD citizen - so that helps a lot here.
IMHO (with my limited NetBSD experience) seems to look bright ... and I wish NetBSD and NetBSD community all the best.
Nice! What a great distribution!
I'm often surprised that the corps didn't pick up the BSDs instead of linux. Given the massively more secret friendly licensing. The exception of course being apple, which forked BSD from NextSTEP.
I'm often surprised that the corps didn't pick up the BSDs instead of linux. Given the massively more secret friendly licensing. The exception of course being apple, which forked BSD from NextSTEP.
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
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
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.
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.
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.
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.
> 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.
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.
https://debugpointnews.com/netbsd-uptime/?amp=1
Some still using BSDs
Sony also uses BSD for the PlayStation 4 and 5 operating systems.
https://en.wikipedia.org/wiki/PlayStation_4_system_software
https://en.wikipedia.org/wiki/PlayStation_4_system_software
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.
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.
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.
It is not a distribution.
It's literally in the name - BSD is "Berkeley software distribution". NetBSD (and Free, Open, Dragon Fly) ship kernels with original tooling and integrated 3rd party software, as a distribution.
Yes it is in the name, much like Federal Express. Not a distribution in a canonical meaning of the word. The idea of distributions is unique to Linux-based systems.
>The idea of distributions is unique to Linux-based systems.
That's a Linux-distribution, and please stop with that nonsense.
https://en.wikipedia.org/wiki/Software_distribution
>>Examples of software distributions include BSD-based distros (such as FreeBSD, NetBSD, OpenBSD, and DragonflyBSD) and Linux-based distros (such as openSUSE, Debian, and Fedora).
That's a Linux-distribution, and please stop with that nonsense.
https://en.wikipedia.org/wiki/Software_distribution
>>Examples of software distributions include BSD-based distros (such as FreeBSD, NetBSD, OpenBSD, and DragonflyBSD) and Linux-based distros (such as openSUSE, Debian, and Fedora).
You should probably eventually try using one of BSD systems, you may like it.
If you look at my past comments (and maybe my Username) you could have seen that FreeBSD is my main-system (Server, Laptop and Workstation) since well over 17 years (the servers since that long).
And yes yes i not just like it but are deeply in love with it ;)
Let us stop here, no hard feeling's and have a good one, cheers.
And yes yes i not just like it but are deeply in love with it ;)
Let us stop here, no hard feeling's and have a good one, cheers.
> The idea of distributions is unique to Linux-based systems.
That's not quite true; there are illumos distros.
That's not quite true; there are illumos distros.
Even if you're right, you should explain your statement when you make it. Otherwise you bait people.
In the meantime, recent and related:
FOSDEM 2024 Schedule - https://news.ycombinator.com/item?id=38694390 - Dec 2023 (76 comments)