proplib (property list library for kernel-userspace communication), RUMP (Rump Kernel - anykernel framework allowing kernel code to run in userspace), Veriexec (verified exec integrity subsystem using cryptographic fingerprints), LFS (Log-structured File System - NetBSD maintains the only production LFS implementation), WAPBL (Write Ahead Physical Block Logging - metadata journaling), ATF (Automated Testing Framework - originated from NetBSD, now used across BSDs), etc. Feel free to check them out! Each one of them is interesting, IMO.
Beyond these specific features, NetBSD's real strength is its extreme portability (runs on over 50 hardware platforms, from VAX to ARM to obscure embedded systems), exceptionally clean and readable codebase, and pkgsrc (a portable package system that works across BSDs, Linux, and other Unix-like systems). I believe it makes NetBSD valuable not just as an OS, but as a reference implementation and research platform. The code quality and documentation are outstanding for anyone studying OS internals as well. :)
I noticed someone asked what NetBSD is. I am happy to break down the different BSD flavors (NetBSD, OpenBSD, FreeBSD, DragonflyBSD) and their unique features if that would be helpful!
I believe NetBSD also holds the record for the fastest kernel boot time for a general purpose operating system -- I held this record with FreeBSD with the system booting in about 23 ms but Emile got NetBSD down to 15 ms IIRC and I've been too busy to work on FreeBSD/Firecracker lately.
I gave a talk about FreeBSD on Firecracker at bsdcan a few years ago. Sorry I'm on my phone just before an orchestra concert so I can't give you a link but if you Google "BSDCan Colin Percival Firecracker" you should find it on YouTube.
I left out FreeBSD from that comment, which has its own set of innovations: Capsicum (capability-based security framework), Jails (OS-level virtualization/containerization which predates Docker by over a decade), MAC Framework (Mandatory Access Control for fine-grained security policies), GEOM (modular disk I/O framework), Linuxulator (Linux binary compatibility layer), ZFS (FreeBSD has arguably the best ZFS implementation outside of Solaris), bhyve (type-2 hypervisor), and so forth.
Userland tools include iocage/bastille (jail managers), poudriere (package building), jemalloc (default allocator which focuses on fragmentation avoidance and scalability) among many others.
Each BSD really does have its own character. FreeBSD leans toward performance and production use, OpenBSD toward security and correctness, NetBSD toward portability and clean design, DragonflyBSD toward alternative SMP approaches!
(illumos/OpenIndiana is quite interesting, too (see DTrace, Doors IPC, Zones, SMF, Contracts, Event Ports, RBAC)).
NetBSD has the neat property that it's possible for a halfway decent C programmer to learn kernel programming from the man pages alone. The documentation is that good. FreeBSD and especially OpenBSD may also share this property, but NetBSD is my go-to OS to tinker with.
Just donated. I run a NetBSD instance on a VPS and on an old Lenovo mini PC that I use as a gateway/file server between my regular network and network of vintage PCs. I have two XT clones, two ATs, some 486s, a Pentium MMX, and an iMac G3 and G4. Fun hobby! I need to get NetBSD running on one of these old machines one of these days.
NetBSD is a powerful force for sustainability. Foundation's commitment to running on a vast array of hardware—new and old—helps reduce e-waste. Old laptops and single-board computers that would otherwise be in a landfill are given new life as robust firewalls, file servers, or even retro-gaming machines, all thanks to NetBSD.
Emotionally I like this - but thinking more dispassionately, these systems use, by modern standards, a huge amount of power. I wonder if, for many (most?) of them, it whould not be more environmentally responsible to replace them with modern, less power-hungry devices.
The cost of creating new computers has got to be pretty high to the environment (I've heard 85% of lifetime carbon emissions from computers are from the manufacturing process), and I strongly suspect that we don't take that into consideration since we greenwash ourselves by forcing China to do our dirty work, chastising them for it, and then patting ourselves on the back for buying "more energy efficient chips".
We need an index of lifecycle costs for products and services, broken down by phase (material production, manufacturing, logistics, operations, disposal, etc.). It's especially needed because those costs are often externalized for buyers (i.e., they aren't built into the price - you don't pay the true cost of gas, for example; many costs are externalized to everyone else). How else can consumers, manufacturers, policymakers, etc. make anything like an informed decision?
I'm surprised someone like the Sierra Club, Consumer Reports, a scientific group, a government group, etc. doesn't undertake it. Yes, it's a bunch of work, there would be uncertainty, but it's essential and better than nothing.
This is why people keep bringing up a carbon tax. It does a lot of the work to internalize all those externalities so individuals or groups don't have to try and figure all this stuff out.
Yes, unfortunately even the best intentioned individuals have very limited ability to make meaningful carbon-minimizing decisions. Carbon tax is such a sensible solution!
They aren't; among other things, most environmental harm is externalized. When you buy things that produce climate change or microplastics, the cost of the impact is paid by society generally.
> The cost of creating new computers has got to be pretty high to the environment
But aren't those made regardless if the people with old computers upgrade to them or not? I guess over time, they'll make less if people buy less, but the ones we'd purchase today has already been made, and might as well replace less energy efficient devices than just being added to the global count.
Manufacturing probably way outshines the usage. And, the side-benefit of using old machines with say NetBSD or OpenBSD etc is that it makes people realize that they don't NEED new modern shiny hardware. I am still daily-driving a x220 on the road (it gets 99% of my stuff done) and at home I am using mostly a slightly upgraded T480. So a 15 year old and a 8 year old laptop. And power consumption wise, my T480 lasts about 17 hours on a charge. That is about same time I'd get out of a modern M3 MacBook Air.
With a bit less performance? Sure. But for my use-case it works. And more people realizing that might actually be a good thing.
T480 can hold 2 batteries. A swappable and an internal one. At least under FreeBSD the swappable alone gives about 10-11 hours on low brightness. The internal one carries the rest of the way.
I did replace the screen with a low power screen, which easily had the biggest impact on battery life.
I haven't tested the T480 under OpenBSD in a while, but my guess is that if I ran it with "apm -L", it would get close to the same numbers.
You're right, the world is better off having some Asian kid taking apart your junk and breathing in solder fumes with no protection, rather than you running a legacy box in your basement that uses some extra kWh per year.
After all, he probably won't live long enough to tell his grandkids stories about power-hungry hardware.
That reminds me, my homelab could use a SPARC box just because.
I get my power off grid through solar generation and battery storage, so I monitor usage pretty closely.
In my experience, newer computers have slightly higher idle power consumption, but much less total consumption for a given compute task. On top of that, new computers are more likely to have dedicated hardware to accelerate the latest codecs.
If you're using an old computer for ray tracing or neural networks or video transcoding, it's probably using enough power that it's worth upgrading.
If you're browsing the Web or watching YouTube videos or running a file server, power consumption is probably similar on old and new computers, but regardless of the age, much higher on desktop computers than laptops.
Look at power supplies on vintage computers, and you'll see that they're much, much smaller than on modern computers.
I’ve just replaced an old Cubieboard (RPi1 alternative - about 2W) running Pi-hole and an old temperamental gigabit router (~10W) with a 2014 Mac mini (plus second Gigabit adapter) - which uses about 11W (a really efficient computer for its age!)
It’s less than the old combo drew - but I wonder if I could be accomplishing the same with an (or even a couple of) SBCs - and if that would ‘pay for itself’ (environmentally well as financially) after a couple of years.
> and if that would ‘pay for itself’ (environmentally well as financially) after a couple of years.
At 11W, the financial costs of running are quite low. I think it's about 90 kWh per year. Depends on your rates, but I've seen $0.60 quoted in comments lately as a high rate for PG&E customers in California, which is about $55. You might well be able to invest in something that can run your load for fewer watts, but I don't know if it's worth the effort.
You're likely to get a lot more savings by looking at things that use more energy.
But how is it draw less power comparing to Raspbian? I’m not arguing, just curious how you measured that, and why NetBSD that much more efficient. I have a RPi2 with Raspbian, and another one with DietPi, I expect them to sip power similarly, considering they’re mostly idle anyway. Would love to read some blog on this though, if you happen to have one.
I too would like to see a side-by-side comparison but if we assume that the claim is true, one reason might be that NetBSD runs very few daemons after a default install. I run NetBSD on a handful of Raspberry Pi Zero machines, and it is really quite a surprise that they run as well as they do for a $5 computer. Your typical Raspbian install has a lot more going on after the default install.
> Your typical Raspbian install has a lot more going on after the default install.
I believe that this is true for just about all widely popular distributions. It's probably possible to set up Arch to have power draw similar to NetBSD, but you're going to have to know what you're doing and it's probably going to require more administration/attention to keep running smoothly than NetBSD does.
That is fascinating, one of the reasons I hadn't done that is a concern about power management features. I'd read openbsd at least had them running flat out most of the time.
It probably works as an economic argument. Readers of this site and many in first world countries have limited barriers to buy new equipment and there are low cost options. Elsewhere that may not be true, either the low cost equipment is still beyond reach, or it is simply difficult to import or completely unavailable.
It's interesting to observe that a 1990 (386) vs 2000 (Pentium III) computer is a heck of a lot different than a 2015 vs 2025 computer. We're talking about Skylake (2015) which would have no issue as a daily driver for quite a lot of people.
It is a logistics issue. People replace perfectly working computers all the time. Sure, if you use Skylake you list performance from Spectre and Meltdown. You may not even be able to run Windows 11 on perfectly working x86-64 machines. But for poor people, power usage may even be more important than for you or me (businesses generally don't care). That Skylake's successors were so terrible is due to Intel and thankfully it allowed AMD and ARM-based SoCs to thrive.
It would be awesome to see a cost breakdown/environmental impact of continuing to run a P4 or G5 today vs. the metals/materials/recycling process/disposal process cost and environmental impact.
I always had the same question about cleaning recycling as it went through a recycling plant -- is the water usage environmentally "friendly" versus what is ultimately recycled (which is often not much, sadly).
That brings up an interesting point. While recycled materials like gold are obviously revenue for the recycler, the cost of polluting water might get externalized in some cases maybe.
In some cases yes, replacing an old machine with a new one can be an evironmentally responsible choice. But in general that's not the case and one should thoughtfully consider the variables including but not limited to software choice, grid carbon cost (see Electricity Maps below), embodied carbon cost of materials, environmental issues of mining and production not strictly related to climate emissions, and more.
Low Tech Magazine wrote an article about this here:
Yeah I don't buy that argument either. The amount of e-waste being saved by NetBSD is so tiny as to be insignificant.
NetBSD is great for retrocomputing, since it's a modern OS that can run on very limited hardware. It's also a very nice traditional UNIX. It's well documented, has a nice codebase, and is a pleasure to use. But for saving e-waste, Linux has it beat.
That old computer running NetBSD probably also isn't running 24/7, making the environmental impact that much lower.
The comparison probably needs to be: Running that old NetBSD machine for a few hours a day, worst case about 40 while I work vs. producing an running a brand new laptop.
If we're talking desktops, then many older machines have 2-300W PSUs, not even enough to power a modern graphics card (I know, an Nvidia card isn't running 600W all the time).
I think it’s a fair question when I’m literally commenting on a statement about sustainability and reuse of old computers.
And on a larger scale, you could use your dismissal against almost everything. Every part of society (and businesses are part of society) can think ‘that thing I do means literally nothing when so much else is going on!’ about almost anything.
There’s a lot of irony in asserting that the ones who are considering and acting upon the consequences of their small choices are the ones exhibiting ‘sheep mentlity’.
Given that there are a lot of cases that old systems are used for that don't run the system full out and that they probably idle a lot when on, just like your new stuff, given that a lot of older hardware barring known stupid designs like Pentium 4 that draw ludicrous power for the perf, and given how much of an absolute shit show electronics recycling is (I've seen a good amount of how it actually gets "broken down"), I honestly think we upgrade way too often for most use cases. If you can still get by doing something important with a P3 or a Core 2, honestly I would be really surprised if it was actually vastly more cost efficient and environmentally friendly to refresh to new hardware.
Well, look, here is my use case. I have some really old hardware that I used to play with earlier in life, it’s located at the parents’ place. I still use that hardware, but not too often, since I live on my own. I could replace all that, as it consumes a lot of power indeed (comparing to SBCs). But in absolute numbers, they’re just regular computers, so power draw is somewhere at 60–90 Watt/hr, I suppose. Which is not that much, especially given I run the machines half a day once a month. What I learned over the years, you can utilise some servers by not running them 24/7. One server there, I wake it over LAN nightly, copy some backups there, and turn it off again. Very useful. So, I think I’d keep them running for as long as they’re alive.
I hope you understand how unique netbsd is, it is one of the only systems which can be compiled so easily with just a single script even from linux or other systems and its rump kernel etc. drivers from what I know are (modular?) so they could be used with other kernels as well if any kernel wants ie.
You never know where the innovation can be, I feel like that each kernel/operating system can bring a new idea, as an example, templeOS uses Holy C which basically is Just in time C (iirc) and that means that you can just edit files of templeOS and restart and those changes would occur
I know TempleOS is niche and a meme OS but I feel like that there are a lot of ideas and unique operating systems and I have heard that netbsd can be good in giving driver support to.
This is just one of many things, and I feel like the main point of NetBSD and the likes are fundamental hackability, they can run on things like routers as well although most run openbsd/freebsd but still. I don't see a reason not to unless you are speaking monetary (ie. it may take some extra funds developing/hosting but that is chump change) but I feel like NETBSD is a novel project with respectable goals and they aren't going to change just for this.
More Options are a good thing. if I can have a project run on Netbsd, then its very easy to port it over to any other vast array of hardware as well, and that hardware includes extremely embedded hardware as well I guess
So in my other comment I mentioned some specific(s) to (or rather, originated from) NetBSD, just as much as for example pledge() (fine-grained system call restriction), unveil() (filesystem visibility restriction), arc4random family[1] (ChaCha20-based CSPRNG), reallocarray() (integer overflow-safe realloc), OpenBGPD (BGP daemon), OpenOSPFD (OSPF daemon), httpd (web server), acme-client (Let's Encrypt client), signify (cryptographic signing tool), etc. are specific to OpenBSD.
DragonflyBSD has some goodies too while we are at it! For example varsym (Variable Symbol System - per-process environment-like variable substitution), nlookup (namecache-based path lookup replacing the vnode-based namei()), objcache (per-CPU object caching allocator), LWKT (Light Weight Kernel Threads - message-passing based threading), HAMMER2 (clustered COW filesystem with multi-master replication, successor to HAMMER), and so forth.
All popular BSDs have their own rich history. I know more about DragonflyBSD than NetBSD, so as an example: DragonflyBSD's core design philosophy centers on SMP scalability (cache-coherent token-based synchronization and LWKT message passing, avoiding fine-grained locking), OpenBSD's gist is security, and so forth.
[1] The ChaCha20-based CSPRNG (originally arc4random was RC4-based), which has been ported to other BSDs and some Linux systems.
(Sorry, I was really tempted to elaborate on these unique features and I felt like your comment was the perfect place for it!).
> I hope you understand how unique netbsd is, it is one of the only systems which can be compiled so easily with just a single script even from linux or other systems and its rump kernel etc. drivers from what I know are (modular?) so they could be used with other kernels as well if any kernel wants
Aren't competing kernels already shipping support for this hardware? Surely the project has to have more selling points than "can be compiled with a single script."
I meant in the sense that since NetBSD supports soooo many devices, it can also help innovation in other kernels if need be as well by being able to take driver support via its rumpkernel as well if need be
And to be honest, I feel like there is this sense of freedom knowing that you can have a system which is portable, if some script can run on my pc on netbsd, chances are if its not too specific, it could run on your pc or even your toaster lol!
Netbsd can run on any device possible and I really appreciate it.
>Surely the project has to have more selling points than "can be compiled with a single script."
Personally I have only heard good things about netbsd but I don't have much expertise in it (sorry), I can recommend you to take a look at smolbsd which looks really cool for uni-kernel purposes as well
> Netbsd can run on any device possible and I really appreciate it.
That's more of a meme than reality and I wish people would actually look into it before mindlessly repeating the trope. I did, when I wanted to run a new OS on a niche device, and the reality is very different. Nowadays Linux works on a lot more hardware than NetBSD does. Yes, NetBSD nominally supports a few more architectures than Linux (very few, especially that μClinux is now upstreamed), but the driver situation for the rest of the system means that it can't run on most devices from those architecture anyway.
This and 'runs' is a very relative term. What exactly works, what doesn't? What do you need? How power efficient or buggy is it? How is the performance? I remember running Linux on an SGI Indy. Not everything worked, as not every Indy was equal (I had six) but also compared to IRIX (especially 5.x series) the performance was dog slow. The Indy had some good hardware (like the soundcard or Indycam) but that wouldn't work (well) on Linux. My Octane 2 didn't have 3D accelerated graphics on Linux. The state of every port isn't equal. Same on OpenBSD. Furthermore, if you run Linux your distribution of choice may not be available or work well.
Note that for most of the more esoteric platforms NetBSD supports, “support” simply means they continue to cross-compile to target that platform. There’s been lots of cases where there’s some major regression that makes a port unusable (can’t boot, can’t interact with the system, etc) and nobody notices for years because there’s both nobody testing these ports and nobody actually using them. At that point, the value of the support is questionable.
> I hope you understand how unique netbsd is, it is one of the only systems which can be compiled so easily with just a single script even from linux or other systems and its rump kernel etc. drivers from what I know are (modular?) so they could be used with other kernels as well if any kernel wants ie.
Linus hast this with User Mode Linux (upstream) and Linux Kernel Library (out of tree).
> You never know where the innovation can be, I feel like that each kernel/operating system can bring a new idea, as an example, templeOS uses Holy C which basically is Just in time C (iirc) and that means that you can just edit files of templeOS and restart and those changes would occur
That's a while ago, but Fabrice Bellard did a demo with his tiny c compiler where it would would compile the Linux Kernel at boot time and then boot the compiled Kernel.
> This is just one of many things, and I feel like the main point of NetBSD and the likes are fundamental hackability, they can run on things like routers as well although most run openbsd/freebsd but still.
Most consumer grade routers run Linux out of the box.
> More Options are a good thing. if I can have a project run on Netbsd, then its very easy to port it over to any other vast array of hardware as well, and that hardware includes extremely embedded hardware as well I guess
uCLinux (upstream) doesn't even need a MMU. It can run on a Cortex-M4 with 8mb ram.
> That's a while ago, but Fabrice Bellard did a demo with his tiny c compiler where it would would compile the Linux Kernel at boot time and then boot the compiled Kernel.
That’s interesting. Do you have a link you can share? Or remember any more details?
Donated! I should have done this months ago when I started using NetBSD for an embedded project idea (that has gone nowhere).
But I feel this link illustrates a big problem with NetBSD’s “no hype” approach: I clicked the link you shared and found an email. The email has the donation link at the very end, and it’s not clickable. When I go to the donation page, there is a ton of text before I even get to see an ugly PayPal tiny button or a tiny form to donate via Stripe.
It’s too hard to notice and too hard to do. The project’s homepage does a better job though. But I think it should be made even more prominent if this is critical for the project’s health!
NetBSD has been a labor of love for a long, long time.
In the mid-90's I was a teenager with a 486-25 on a desk in a closet running NetBSD 0.9-1.0, connected to 10base2 going to my dad's office where there was a computer that dual booted to Linux. I learned so much from those systems; systems programming, how to really use the C programming language, sysadmin skills, reading network traces. A whole part of who I am today derives from those early experiences trying to figure out what the $## was going on while tracking -CURRENT.
I'm retired from tech and a high school teacher these days and allowed to teach wack/out of level things.
I would love to teach operating systems with NetBSD, but between the space hardware stuff I do and the Verilog/digital logic/microprocessor architecture class I teach, I soak up all the interested students' elective slots.
Same here. But one other thing to add for new responses about "Why NetBSD", the rump kernel.
Years ago I had to get a very old document off of a DOS diskette. So I tried:
* On Linux: accessing the diskette would cause a panic or a reboot or massive read failures.
* FreeBSD: panics all the time
* NetBSD: panics. But then I remembered it had rump. So I said, why not try that. Started up rump, got a few code dumps, but after a some tries I got a bit over 90% of the document off of the diskette. The main system had no issues with the rump kernel crashing.
So that alone is worth the "price of admission" :)
Donated. I think what NetBSD is missing, but also hard to pull off, is something like https://openbsd.amsterdam. A service that people can use where parts of the fees go towards the foundation.
Out of pure curiosity, is all the actual programming work for the foundation provided free of charge by volunteers? And the foundation expenses are mostly legal and administrative?
>Maybe the consulting section includes payments for programming work? Presumably at cheap rates, if so?
Have we reached the point on the timeline where we believe low-level operating system code should be acquired at "cheap rates"? While simultaneously, I assume, believing webshit cloud bollocks still demands top dollar?
I do not hold that belief, and nothing in my reply implies that I do.
Per the link I included, the total spent on consulting was 17,939.51 USD. So, if they were paying people, the people were working cheaply. But the consulting may have been for non-programming work. Hence my question.
Donated!! I run a small cluster of a few nodes I bought for cheap, and I’m experimenting with SSI on them. The kernel is really nice to read and modify.
Donated! Thank you very much, NetBSD was one of my first experiences, on a Pentium 60 with a 504MB hard drive. It made me who I am today, eternally grateful to have learned from such amazing and talented people.
Do they offer a swag store like OpenBSD or FreeBSD? I realize they only get pennies from those sales but that’s typically my approach, buy a shirt for $30 and make an extra $20 donation.
Big fan of NetBSD. From the perspective of a kernel hacker, I found the NetBSD codebase to be very readable and easy to work with. FreeBSD and OpenBSD, while being more fully-featured and security-minded respectively, had to make compromises in their codebase as a result.
I just donated. It's important to keep projects like NetBSD vital, as a monoculture benefits nobody but the monoculturists. I think of it as a way to help ensure Survival of the Species - by diversity.
It's really in the best interest of everyone using it to chip in and keep the project relevant. Unfortunately the amount of donations is going to be contingent on the size of its user base which will need to grow to ensure its longevity.
Maybe not yet but I can see Linux's place as the shitbox saviour start slipping a bit in the next few years. Debian dropping x86, distros getting fatter in general.. I can't really see those trends reversing. Meanwhile NetBSD goes against them.
However it goes, the main issue is one no operating system can solve which is modern life relying on the Web and beefier browsers. Unless you want to rebel against that you're probably better off getting a laptop from the past 10 years for < £100 on eBay.
Although I agree with beefier browsers, I also want to say that there are browsers like dillo etc. which can be good enough for simple websites and also not everything needs a web browser to be usable
Imagine this, a system which can watch movies, edit texts, create disks, have curl/wget, send and recieve files using piping server (which is a curl thing) , view pdfs, mpv and what not, a desktop manager, file manager etc.
As someone hacking around with the legendary tiny core linux, I am more and more mind blown each day with just how much can happen in 14-21 MB, you can definitely build a mini self hosting rack with just some remastering as tinycore can actually run podman as well (combine this with alpine containers to create a super duper minimalist self hosting things too)
the possibilities are endless. When I ran tiny core linux on my pc and ran nothing else, It took 21 mb in ram for a whole gui with editors and file managers etc. all running in ram so super fast filesystem with a package manager
I personally wanted to build my own operating system to limit myself to the most minimal system so taht I just study and do nothing else, I thought tiny core was it but then I tried to hack around it and there are sooooo many things in 21 mb, makes me appreciate minimalism
I have to say, the sheer fucking irony of this statement made me do a double-take.
I might be showing my age a bit, but I'm still remembering when web-browsing was considered a "light" activity (without extensions like Web Java), and watching a video was "very computationally expensive".
I guess some shift happened in the early 2010's where video playback was hardware accelerated more frequently; and complicated javascript started taking off as Google unveiled v8.
As much as the tech bro billionare class salivate over the idea of bringing an "everything app" to the market, we already have those. They're called browsers.
How is TLS negotiation and transport on older hardware (with no AES-NI hardware acceleration)?
I remember it used to be expensive as heck to do TLS back in 2014~, so much so that we bought accelerator cards and segmented "secure" servers so that load wouldn't hit the ordinary browsing of our sites...
The argument for NetBSD is that it runs on almost anything that was ever produced. That isn't the case for Linux, even older x86 is no longer supported in the mainline.
Linux (the kernel) may have been ported to more machines and architectures than NetBSD’s kernel, yes. But is all the code present in the same source tree or do you have to go find patch sets or unofficial branches?
More importantly: is there a modern distribution that builds an installable system for that platform?
The special thing about NetBSD is that you get the portability out of a single and modern tree for many more platforms than any single Linux distribution offers.
What do you think, there are milions of people or companies running NetBSD on 486 to protect the planet from e-waste? How many times have you replaced your phone with a newer model in the last 10 years?
An interesting thread about the upgrade exists on the same list that the link points to, just a few messages earlier. Hopefully, with the donations, the project can improve the existing friction points - https://mail-index.netbsd.org/netbsd-users/2025/10/01/msg033...
Things like NetBSD seem like an obvious use case for tokens and DAOs (funding and governing treasuries for non-crypto open source software projects). Why is this not more common already?
“Then there is the % address operator: user %domainB@domainA is first sent to domainA, which expands the rightmost (in this case, the only) percent sign to an @ sign. The address is now user@domainB, and the mailer happily forwards your message to domainB, which delivers it to user. This type of address is sometimes referred to as “Ye Olde ARPAnet Kludge,” and its use is discouraged“
I would guess it's an anti-spam measure. Although if I'm reading sibling comment right, it is actually a valid email address? (Assuming you have a mail server running on localhost.)
This year I've seen some retro tech YouTube videos about people putting modern NetBSD in their expensive PDPs and Vax machines. Dave Plumber comes to mind.
Donated. While I don't use NetBSD, the existence of projects like this is essential for open source technology, operating system design, and the overall vitality of programming culture.
It's my nameserver/DHCP server. I used to have it set up as an iSCSI target for backups and as a boot server for my firewall, but I do something else these days.
My main reason for using NetBSD for this is to have easy access to the man pages. Like the other BSDs, the man pages are exceptionally well-written and are a tremendous resource for doing POSIX programming. Plus I find myself digging through the code when I'm interested in how something is implemented. Having a local repository of good C code with a liberal license is worth having the extra OS to manage.
You aren't going to see OpenBSD share a kernel with anyone - it's too different and makes trade-offs the others won't accept. And NetBSD doesn't need the heavyweight kernel FreeBSD uses.
From what I've seen, the BSD community swaps code around on a regular basis. But they pick and choose what code to use based on their own goals. It seems to work pretty well.
NetBSD is the slim, small, traditional BSD that has an emphasis on clean code and portability. It's great for small jobs and it'll run on that old SPARC that's collecting dust in the closet. It's simpler than FreeBSD (the industrial strength BSD) and doesn't have the hyper focus on security that OpenBSD does.
Absent something like OpenSolaris really taking off, the popular opinion in my circles is that *BSD would have (which is at least related to the same thing). Unless you believe Windows would have "won" which is certainly the side-bet a lot of companies were making at the time.
Not really, as anyone that was already using UNIX at the time is aware of, and the whole UNIX V6 vs BSD integrations as baseline for Aix, Solaris, Irix, HP-UX,....
No, we wouldn't. Linux climbed its way up to overtake proprietary UNIX despite being less capable, which it very much was at the time.
Linux came around at the right time when the Internet was going public and regular people had access to hardware that could run a decent UNIX. People latched onto it because it was free and an interesting project. The free BSDs were just late enough to the party that they missed out on the momentum.
All the proprietary UNIX vendors (other than SCO) relied on expensive proprietary hardware sales. Intel ate their lunch while they were too busy stabbing each other in the back to notice. Linux killed SCO because SCO was, quite frankly, overpriced crap.
None of this had anything to do with the license, other that the fact you could use it for free. It was all about hardware availability, the rise of the Internet, the wave of new IT people who had experienced Linux at home, and the fact that Linux on Intel was good enough to replace those pricy proprietary machines.
Now, you wanna talk Apple, there's where your code "theft" kicks in. But that's a whole different thing.
I wonder if there'll be some big cultural shift in open source as people get more annoyed at cases of big companies taking their code and giving nothing back/demanding them to work for free. Might already be happening just slowly
If you don't work for that big company, you're more than welcome to ignore any "demands" to work for free. And a lot of people in open source are indeed paid to work on it.
AGPL and non-compete licenses like SSPL and FSL have been steadily adopted —Redis, ElasticSearch and Liquibase notably switched to them—, pretty much motivated by not wanting to be screwed over by Big Tech one way or another.
More than fine if you ask me, giving away your work for megacorps and oligarchs to steamroll your business or otherwise society at large isn’t much of a public service in the end
Yeah I have been thinking more and more about source available. I am not sure about AGPL as I feel like, some companies might still use it. honestly, it depends but all the companies which use non compete license somehow diverts back to AGPL if they were open source before
So if some company/product was open source and then used source available license, the backlash would be so much that they go to something like AGPl most of the time
but that happens because people feel betrayed because some might have contributed thinking its foss forever so its a rug pull
I think a good idea could be to have a source available license from the start so that everybody who ever contributes knows this as a fact.
What are your thoughts? What should I or anyone else pick? As a "foss" advocate, I would prefer AGPL but I don't want to get screwed by Big Tech ever with all the loopholes that they can have (like AWS), Honestly I don't know which is why I am asking really.
My personal (non-mainstream) thought on this topic is: work on open source projects that serve a purpose that is very antithetical to the interests of bigtech companies. This way, such companies will be a lot less interested in "using" your project (without contributing back).
From Big Tech to startups, most corporations waste billions in ridiculous and frivolous projects and yet insignificant money is sent back to the projects they owe their very existence.
But don't you dare switch to a proprietary license or you will be dragged across social media as an evil selfish person. Even if it's only postponing source releases for a couple of years.
"BSD" doesn't really exist as far as I'm aware. It was a proprietary operating system made at Berkeley for studying OS design.
OpenBSD/FreeBSD/NetBSD are open source continuations of the source-available 4.4-BSDlite code (removing AT&T proprietary extensions iirc).
OpenBSD follows BSD principles but focuses on code clarity and security.
FreeBSD tries to be very flexible, putting user-experience over security. (it has to be noted that OpenBSD is very usable, but lacks a lot of nice features like ZFS and DTrace that FreeBSD supports).
NetBSD is all about being incredibly lean and portable. NetBSD will run on basically anything, even things that Linux and other *BSD's have no hope of running on.
There absolutely was a real-world non-research operating system called just "BSD", and it ran primarily on DEC VAX hardware to start. Other Unixes like SunOS eventually started basing themselves on it. The last official Berkeley Software Distribution was 4.4BSD-Lite in 1994, which by then had been ported to many other platforms, including i386. The Computer Systems Research Group which maintained BSD disbanded the next year.
NetBSD has some goodies, too, by the way!
proplib (property list library for kernel-userspace communication), RUMP (Rump Kernel - anykernel framework allowing kernel code to run in userspace), Veriexec (verified exec integrity subsystem using cryptographic fingerprints), LFS (Log-structured File System - NetBSD maintains the only production LFS implementation), WAPBL (Write Ahead Physical Block Logging - metadata journaling), ATF (Automated Testing Framework - originated from NetBSD, now used across BSDs), etc. Feel free to check them out! Each one of them is interesting, IMO.
Beyond these specific features, NetBSD's real strength is its extreme portability (runs on over 50 hardware platforms, from VAX to ARM to obscure embedded systems), exceptionally clean and readable codebase, and pkgsrc (a portable package system that works across BSDs, Linux, and other Unix-like systems). I believe it makes NetBSD valuable not just as an OS, but as a reference implementation and research platform. The code quality and documentation are outstanding for anyone studying OS internals as well. :)
I noticed someone asked what NetBSD is. I am happy to break down the different BSD flavors (NetBSD, OpenBSD, FreeBSD, DragonflyBSD) and their unique features if that would be helpful!
I believe NetBSD also holds the record for the fastest kernel boot time for a general purpose operating system -- I held this record with FreeBSD with the system booting in about 23 ms but Emile got NetBSD down to 15 ms IIRC and I've been too busy to work on FreeBSD/Firecracker lately.
Interesting! Both 23ms and 15ms are lightning fast. Have you written about it anywhere in more detail?
https://smolbsd.org/
I gave a talk about FreeBSD on Firecracker at bsdcan a few years ago. Sorry I'm on my phone just before an orchestra concert so I can't give you a link but if you Google "BSDCan Colin Percival Firecracker" you should find it on YouTube.
https://www.youtube.com/watch?v=MT3cdeuRTzs - Porting FreeBSD to Firecracker by:Colin Percival?
Yup that's the one.
Please do break down the differences.
I could not hold myself back, so I already elaborated on OpenBSD and DragonflyBSD here (albeit it is non-exhaustive): https://news.ycombinator.com/item?id=45714491 :D
I left out FreeBSD from that comment, which has its own set of innovations: Capsicum (capability-based security framework), Jails (OS-level virtualization/containerization which predates Docker by over a decade), MAC Framework (Mandatory Access Control for fine-grained security policies), GEOM (modular disk I/O framework), Linuxulator (Linux binary compatibility layer), ZFS (FreeBSD has arguably the best ZFS implementation outside of Solaris), bhyve (type-2 hypervisor), and so forth.
Userland tools include iocage/bastille (jail managers), poudriere (package building), jemalloc (default allocator which focuses on fragmentation avoidance and scalability) among many others.
Each BSD really does have its own character. FreeBSD leans toward performance and production use, OpenBSD toward security and correctness, NetBSD toward portability and clean design, DragonflyBSD toward alternative SMP approaches!
(illumos/OpenIndiana is quite interesting, too (see DTrace, Doors IPC, Zones, SMF, Contracts, Event Ports, RBAC)).
NetBSD has the neat property that it's possible for a halfway decent C programmer to learn kernel programming from the man pages alone. The documentation is that good. FreeBSD and especially OpenBSD may also share this property, but NetBSD is my go-to OS to tinker with.
For sure! I agree.
OpenBSD has great man pages as well. :)
Just donated. I run a NetBSD instance on a VPS and on an old Lenovo mini PC that I use as a gateway/file server between my regular network and network of vintage PCs. I have two XT clones, two ATs, some 486s, a Pentium MMX, and an iMac G3 and G4. Fun hobby! I need to get NetBSD running on one of these old machines one of these days.
Thanks a lot :)
Do any governments use NetBSD? If so, they should be the #1 name and shame.
I guess we're all complicit to <insert war crime> for using and developing software that anyone can use freely...
Yes, let's name and shame software users /s
”Now, that, I got me some Seagram's gin, Everybody got they cups, but they ain't chipped in”—Snoop
NetBSD is a powerful force for sustainability. Foundation's commitment to running on a vast array of hardware—new and old—helps reduce e-waste. Old laptops and single-board computers that would otherwise be in a landfill are given new life as robust firewalls, file servers, or even retro-gaming machines, all thanks to NetBSD.
Emotionally I like this - but thinking more dispassionately, these systems use, by modern standards, a huge amount of power. I wonder if, for many (most?) of them, it whould not be more environmentally responsible to replace them with modern, less power-hungry devices.
You know, I wonder about that.
The cost of creating new computers has got to be pretty high to the environment (I've heard 85% of lifetime carbon emissions from computers are from the manufacturing process), and I strongly suspect that we don't take that into consideration since we greenwash ourselves by forcing China to do our dirty work, chastising them for it, and then patting ourselves on the back for buying "more energy efficient chips".
We need an index of lifecycle costs for products and services, broken down by phase (material production, manufacturing, logistics, operations, disposal, etc.). It's especially needed because those costs are often externalized for buyers (i.e., they aren't built into the price - you don't pay the true cost of gas, for example; many costs are externalized to everyone else). How else can consumers, manufacturers, policymakers, etc. make anything like an informed decision?
I'm surprised someone like the Sierra Club, Consumer Reports, a scientific group, a government group, etc. doesn't undertake it. Yes, it's a bunch of work, there would be uncertainty, but it's essential and better than nothing.
There are such things for food:
* Klimato: https://klimato.com/
* The Big Climate Database: http://thebigclimatedatabase.com/
This is why people keep bringing up a carbon tax. It does a lot of the work to internalize all those externalities so individuals or groups don't have to try and figure all this stuff out.
Yes, unfortunately even the best intentioned individuals have very limited ability to make meaningful carbon-minimizing decisions. Carbon tax is such a sensible solution!
Money dollars are a pretty good “stand-in” but in general it’s better to reuse and reduce than to recycle.
But if the cost of a new machine is the same as a year or two of the old operating the new is probably the way to go.
> Money dollars are a pretty good “stand-in”
They aren't; among other things, most environmental harm is externalized. When you buy things that produce climate change or microplastics, the cost of the impact is paid by society generally.
I'm saying that I think "cost" in terms of money is not a good proxy for environmental impact.
> The cost of creating new computers has got to be pretty high to the environment
But aren't those made regardless if the people with old computers upgrade to them or not? I guess over time, they'll make less if people buy less, but the ones we'd purchase today has already been made, and might as well replace less energy efficient devices than just being added to the global count.
I think you answered your own question here.
Manufacturing probably way outshines the usage. And, the side-benefit of using old machines with say NetBSD or OpenBSD etc is that it makes people realize that they don't NEED new modern shiny hardware. I am still daily-driving a x220 on the road (it gets 99% of my stuff done) and at home I am using mostly a slightly upgraded T480. So a 15 year old and a 8 year old laptop. And power consumption wise, my T480 lasts about 17 hours on a charge. That is about same time I'd get out of a modern M3 MacBook Air.
With a bit less performance? Sure. But for my use-case it works. And more people realizing that might actually be a good thing.
17 hours?! I can’t believe it! How did you optimise the machine to be that energy efficient?
T480 can hold 2 batteries. A swappable and an internal one. At least under FreeBSD the swappable alone gives about 10-11 hours on low brightness. The internal one carries the rest of the way.
I did replace the screen with a low power screen, which easily had the biggest impact on battery life.
I haven't tested the T480 under OpenBSD in a while, but my guess is that if I ran it with "apm -L", it would get close to the same numbers.
Oh wow, thanks! And a low power screen is what? I thought they’re mostly the same, with probably exception of the most recent oled ones.
Outside of keeping a SPARC blade system running 24/7, most old systems use similar power to a light fixture.
This argument misses the forest for the trees for non-commercial users.
You're right, the world is better off having some Asian kid taking apart your junk and breathing in solder fumes with no protection, rather than you running a legacy box in your basement that uses some extra kWh per year.
After all, he probably won't live long enough to tell his grandkids stories about power-hungry hardware.
That reminds me, my homelab could use a SPARC box just because.
I get my power off grid through solar generation and battery storage, so I monitor usage pretty closely.
In my experience, newer computers have slightly higher idle power consumption, but much less total consumption for a given compute task. On top of that, new computers are more likely to have dedicated hardware to accelerate the latest codecs.
If you're using an old computer for ray tracing or neural networks or video transcoding, it's probably using enough power that it's worth upgrading.
If you're browsing the Web or watching YouTube videos or running a file server, power consumption is probably similar on old and new computers, but regardless of the age, much higher on desktop computers than laptops.
Look at power supplies on vintage computers, and you'll see that they're much, much smaller than on modern computers.
I have a raspberry Pi 2 running NetBSD at home providing a lot of useful services.
It uses just 1.5W.
Any more recent alternative would consume much more power.
That’s sort of why this is on my mind.
I’ve just replaced an old Cubieboard (RPi1 alternative - about 2W) running Pi-hole and an old temperamental gigabit router (~10W) with a 2014 Mac mini (plus second Gigabit adapter) - which uses about 11W (a really efficient computer for its age!)
It’s less than the old combo drew - but I wonder if I could be accomplishing the same with an (or even a couple of) SBCs - and if that would ‘pay for itself’ (environmentally well as financially) after a couple of years.
> and if that would ‘pay for itself’ (environmentally well as financially) after a couple of years.
At 11W, the financial costs of running are quite low. I think it's about 90 kWh per year. Depends on your rates, but I've seen $0.60 quoted in comments lately as a high rate for PG&E customers in California, which is about $55. You might well be able to invest in something that can run your load for fewer watts, but I don't know if it's worth the effort.
You're likely to get a lot more savings by looking at things that use more energy.
Why is it NetBSD and which services do you have? I’d love to read about that, and I think I might utilise mine that way too.
NetBSD because I love simplicity and because it draws less power than running Raspbian.
It is running Gitea, Prometheus and a bunch of small tools I wrote (weather monitoring, home automation, family data sharing and tools for my Kids)
But how is it draw less power comparing to Raspbian? I’m not arguing, just curious how you measured that, and why NetBSD that much more efficient. I have a RPi2 with Raspbian, and another one with DietPi, I expect them to sip power similarly, considering they’re mostly idle anyway. Would love to read some blog on this though, if you happen to have one.
I too would like to see a side-by-side comparison but if we assume that the claim is true, one reason might be that NetBSD runs very few daemons after a default install. I run NetBSD on a handful of Raspberry Pi Zero machines, and it is really quite a surprise that they run as well as they do for a $5 computer. Your typical Raspbian install has a lot more going on after the default install.
> Your typical Raspbian install has a lot more going on after the default install.
I believe that this is true for just about all widely popular distributions. It's probably possible to set up Arch to have power draw similar to NetBSD, but you're going to have to know what you're doing and it's probably going to require more administration/attention to keep running smoothly than NetBSD does.
I just measured it by myself with a domestic power usage tool.
That is fascinating, one of the reasons I hadn't done that is a concern about power management features. I'd read openbsd at least had them running flat out most of the time.
It probably works as an economic argument. Readers of this site and many in first world countries have limited barriers to buy new equipment and there are low cost options. Elsewhere that may not be true, either the low cost equipment is still beyond reach, or it is simply difficult to import or completely unavailable.
It's interesting to observe that a 1990 (386) vs 2000 (Pentium III) computer is a heck of a lot different than a 2015 vs 2025 computer. We're talking about Skylake (2015) which would have no issue as a daily driver for quite a lot of people.
It is a logistics issue. People replace perfectly working computers all the time. Sure, if you use Skylake you list performance from Spectre and Meltdown. You may not even be able to run Windows 11 on perfectly working x86-64 machines. But for poor people, power usage may even be more important than for you or me (businesses generally don't care). That Skylake's successors were so terrible is due to Intel and thankfully it allowed AMD and ARM-based SoCs to thrive.
It would be awesome to see a cost breakdown/environmental impact of continuing to run a P4 or G5 today vs. the metals/materials/recycling process/disposal process cost and environmental impact.
I always had the same question about cleaning recycling as it went through a recycling plant -- is the water usage environmentally "friendly" versus what is ultimately recycled (which is often not much, sadly).
That brings up an interesting point. While recycled materials like gold are obviously revenue for the recycler, the cost of polluting water might get externalized in some cases maybe.
In some cases yes, replacing an old machine with a new one can be an evironmentally responsible choice. But in general that's not the case and one should thoughtfully consider the variables including but not limited to software choice, grid carbon cost (see Electricity Maps below), embodied carbon cost of materials, environmental issues of mining and production not strictly related to climate emissions, and more.
Low Tech Magazine wrote an article about this here:
https://solar.lowtechmagazine.com/2020/12/how-and-why-i-stop...
https://app.electricitymaps.com
Yeah I don't buy that argument either. The amount of e-waste being saved by NetBSD is so tiny as to be insignificant.
NetBSD is great for retrocomputing, since it's a modern OS that can run on very limited hardware. It's also a very nice traditional UNIX. It's well documented, has a nice codebase, and is a pleasure to use. But for saving e-waste, Linux has it beat.
This was the reason I moved from my sturdy ~2008 laptop, power efficiency. That said I don't buy new anyway, i just jumped to the 2010-2015 range.
Always funny to read from environmentalists discussing how much power draws a lenovo laptop from 2014 running NetBSD.
Come on, look at all the businesses and what's really happening in the space you're commenting on. That laptop literally means nothing.
Even the Pentium M systems from a decade prior have <10W idle. Honestly embarrassing to mention when AI chips are dumping kilowatts.
Wonder how long you could run an old laptop on the energy used by a single ChatGPT conversation?
Wonder how useful the ChatGPT conversation is compared to a Google search from begin '00s (when Google was new and something like Pentium 4 was new).
Months at the minimum probably.
That old computer running NetBSD probably also isn't running 24/7, making the environmental impact that much lower.
The comparison probably needs to be: Running that old NetBSD machine for a few hours a day, worst case about 40 while I work vs. producing an running a brand new laptop.
If we're talking desktops, then many older machines have 2-300W PSUs, not even enough to power a modern graphics card (I know, an Nvidia card isn't running 600W all the time).
I think it’s a fair question when I’m literally commenting on a statement about sustainability and reuse of old computers.
And on a larger scale, you could use your dismissal against almost everything. Every part of society (and businesses are part of society) can think ‘that thing I do means literally nothing when so much else is going on!’ about almost anything.
Be passive, do nothing. Not even asking. Just consume. Your actions mean nothing because whatabout.
Lol. You're not using your laptop making huge impact. Wake up, be real. Sheep mentality, exactly what they want you to believe in.
There’s a lot of irony in asserting that the ones who are considering and acting upon the consequences of their small choices are the ones exhibiting ‘sheep mentlity’.
Also, all of the devices that could run NetBSD can just run Linux
Given that there are a lot of cases that old systems are used for that don't run the system full out and that they probably idle a lot when on, just like your new stuff, given that a lot of older hardware barring known stupid designs like Pentium 4 that draw ludicrous power for the perf, and given how much of an absolute shit show electronics recycling is (I've seen a good amount of how it actually gets "broken down"), I honestly think we upgrade way too often for most use cases. If you can still get by doing something important with a P3 or a Core 2, honestly I would be really surprised if it was actually vastly more cost efficient and environmentally friendly to refresh to new hardware.
Computers of that vintage drew way more power at idle than modern systems.
most older systems are surprisingly frugal on idle and are typically used sporadically
Well, look, here is my use case. I have some really old hardware that I used to play with earlier in life, it’s located at the parents’ place. I still use that hardware, but not too often, since I live on my own. I could replace all that, as it consumes a lot of power indeed (comparing to SBCs). But in absolute numbers, they’re just regular computers, so power draw is somewhere at 60–90 Watt/hr, I suppose. Which is not that much, especially given I run the machines half a day once a month. What I learned over the years, you can utilise some servers by not running them 24/7. One server there, I wake it over LAN nightly, copy some backups there, and turn it off again. Very useful. So, I think I’d keep them running for as long as they’re alive.
Also I bet >50% of personal computer e waste is bog standard x86-64 by now. No need to support a vast array of hardware.
> No need to support a vast array of hardware
I hope you understand how unique netbsd is, it is one of the only systems which can be compiled so easily with just a single script even from linux or other systems and its rump kernel etc. drivers from what I know are (modular?) so they could be used with other kernels as well if any kernel wants ie.
You never know where the innovation can be, I feel like that each kernel/operating system can bring a new idea, as an example, templeOS uses Holy C which basically is Just in time C (iirc) and that means that you can just edit files of templeOS and restart and those changes would occur
I know TempleOS is niche and a meme OS but I feel like that there are a lot of ideas and unique operating systems and I have heard that netbsd can be good in giving driver support to.
This is just one of many things, and I feel like the main point of NetBSD and the likes are fundamental hackability, they can run on things like routers as well although most run openbsd/freebsd but still. I don't see a reason not to unless you are speaking monetary (ie. it may take some extra funds developing/hosting but that is chump change) but I feel like NETBSD is a novel project with respectable goals and they aren't going to change just for this.
More Options are a good thing. if I can have a project run on Netbsd, then its very easy to port it over to any other vast array of hardware as well, and that hardware includes extremely embedded hardware as well I guess
I agree.
So in my other comment I mentioned some specific(s) to (or rather, originated from) NetBSD, just as much as for example pledge() (fine-grained system call restriction), unveil() (filesystem visibility restriction), arc4random family[1] (ChaCha20-based CSPRNG), reallocarray() (integer overflow-safe realloc), OpenBGPD (BGP daemon), OpenOSPFD (OSPF daemon), httpd (web server), acme-client (Let's Encrypt client), signify (cryptographic signing tool), etc. are specific to OpenBSD.
DragonflyBSD has some goodies too while we are at it! For example varsym (Variable Symbol System - per-process environment-like variable substitution), nlookup (namecache-based path lookup replacing the vnode-based namei()), objcache (per-CPU object caching allocator), LWKT (Light Weight Kernel Threads - message-passing based threading), HAMMER2 (clustered COW filesystem with multi-master replication, successor to HAMMER), and so forth.
All popular BSDs have their own rich history. I know more about DragonflyBSD than NetBSD, so as an example: DragonflyBSD's core design philosophy centers on SMP scalability (cache-coherent token-based synchronization and LWKT message passing, avoiding fine-grained locking), OpenBSD's gist is security, and so forth.
[1] The ChaCha20-based CSPRNG (originally arc4random was RC4-based), which has been ported to other BSDs and some Linux systems.
(Sorry, I was really tempted to elaborate on these unique features and I felt like your comment was the perfect place for it!).
> I hope you understand how unique netbsd is, it is one of the only systems which can be compiled so easily with just a single script even from linux or other systems and its rump kernel etc. drivers from what I know are (modular?) so they could be used with other kernels as well if any kernel wants
Aren't competing kernels already shipping support for this hardware? Surely the project has to have more selling points than "can be compiled with a single script."
Support for x86_64?
I meant in the sense that since NetBSD supports soooo many devices, it can also help innovation in other kernels if need be as well by being able to take driver support via its rumpkernel as well if need be
And to be honest, I feel like there is this sense of freedom knowing that you can have a system which is portable, if some script can run on my pc on netbsd, chances are if its not too specific, it could run on your pc or even your toaster lol!
https://laughingsquid.com/netbsd-toaster/
Netbsd can run on any device possible and I really appreciate it.
>Surely the project has to have more selling points than "can be compiled with a single script."
Personally I have only heard good things about netbsd but I don't have much expertise in it (sorry), I can recommend you to take a look at smolbsd which looks really cool for uni-kernel purposes as well
https://news.ycombinator.com/item?id=45582758
I feel like that there is a lot more things that can be done with netbsd as well or other open source projects in general as well
> Netbsd can run on any device possible and I really appreciate it.
That's more of a meme than reality and I wish people would actually look into it before mindlessly repeating the trope. I did, when I wanted to run a new OS on a niche device, and the reality is very different. Nowadays Linux works on a lot more hardware than NetBSD does. Yes, NetBSD nominally supports a few more architectures than Linux (very few, especially that μClinux is now upstreamed), but the driver situation for the rest of the system means that it can't run on most devices from those architecture anyway.
This and 'runs' is a very relative term. What exactly works, what doesn't? What do you need? How power efficient or buggy is it? How is the performance? I remember running Linux on an SGI Indy. Not everything worked, as not every Indy was equal (I had six) but also compared to IRIX (especially 5.x series) the performance was dog slow. The Indy had some good hardware (like the soundcard or Indycam) but that wouldn't work (well) on Linux. My Octane 2 didn't have 3D accelerated graphics on Linux. The state of every port isn't equal. Same on OpenBSD. Furthermore, if you run Linux your distribution of choice may not be available or work well.
Note that for most of the more esoteric platforms NetBSD supports, “support” simply means they continue to cross-compile to target that platform. There’s been lots of cases where there’s some major regression that makes a port unusable (can’t boot, can’t interact with the system, etc) and nobody notices for years because there’s both nobody testing these ports and nobody actually using them. At that point, the value of the support is questionable.
> I hope you understand how unique netbsd is, it is one of the only systems which can be compiled so easily with just a single script even from linux or other systems and its rump kernel etc. drivers from what I know are (modular?) so they could be used with other kernels as well if any kernel wants ie.
Linus hast this with User Mode Linux (upstream) and Linux Kernel Library (out of tree).
> You never know where the innovation can be, I feel like that each kernel/operating system can bring a new idea, as an example, templeOS uses Holy C which basically is Just in time C (iirc) and that means that you can just edit files of templeOS and restart and those changes would occur
That's a while ago, but Fabrice Bellard did a demo with his tiny c compiler where it would would compile the Linux Kernel at boot time and then boot the compiled Kernel.
> This is just one of many things, and I feel like the main point of NetBSD and the likes are fundamental hackability, they can run on things like routers as well although most run openbsd/freebsd but still.
Most consumer grade routers run Linux out of the box.
> More Options are a good thing. if I can have a project run on Netbsd, then its very easy to port it over to any other vast array of hardware as well, and that hardware includes extremely embedded hardware as well I guess
uCLinux (upstream) doesn't even need a MMU. It can run on a Cortex-M4 with 8mb ram.
> That's a while ago, but Fabrice Bellard did a demo with his tiny c compiler where it would would compile the Linux Kernel at boot time and then boot the compiled Kernel.
That’s interesting. Do you have a link you can share? Or remember any more details?
I’m curious how long it took to fully start
Have you considered any advantages with ensuring code is portable?
Isn’t that an advantage by itself?
Donated! I should have done this months ago when I started using NetBSD for an embedded project idea (that has gone nowhere).
But I feel this link illustrates a big problem with NetBSD’s “no hype” approach: I clicked the link you shared and found an email. The email has the donation link at the very end, and it’s not clickable. When I go to the donation page, there is a ton of text before I even get to see an ugly PayPal tiny button or a tiny form to donate via Stripe.
It’s too hard to notice and too hard to do. The project’s homepage does a better job though. But I think it should be made even more prominent if this is critical for the project’s health!
https://netbsd.org/donations/
Donated. I hope NetBSD becomes a stronger option for my old PCs. So many good old machines that could benefit from it.
Thanks a lot for donations:)
Thank -you- for all you do! Donated.
NetBSD has been a labor of love for a long, long time.
In the mid-90's I was a teenager with a 486-25 on a desk in a closet running NetBSD 0.9-1.0, connected to 10base2 going to my dad's office where there was a computer that dual booted to Linux. I learned so much from those systems; systems programming, how to really use the C programming language, sysadmin skills, reading network traces. A whole part of who I am today derives from those early experiences trying to figure out what the $## was going on while tracking -CURRENT.
Thanks a lot :)
Here something you might like : CS631 -- Advanced Programming in the UNIX Environment - NetBSD https://stevens.netmeister.org/631/netbsd.html
:D I've actually seen it.
I'm retired from tech and a high school teacher these days and allowed to teach wack/out of level things.
I would love to teach operating systems with NetBSD, but between the space hardware stuff I do and the Verilog/digital logic/microprocessor architecture class I teach, I soak up all the interested students' elective slots.
Nice.. :)
Same here. But one other thing to add for new responses about "Why NetBSD", the rump kernel.
Years ago I had to get a very old document off of a DOS diskette. So I tried:
* On Linux: accessing the diskette would cause a panic or a reboot or massive read failures.
* FreeBSD: panics all the time
* NetBSD: panics. But then I remembered it had rump. So I said, why not try that. Started up rump, got a few code dumps, but after a some tries I got a bit over 90% of the document off of the diskette. The main system had no issues with the rump kernel crashing.
So that alone is worth the "price of admission" :)
Wow this is cool. I always get fascinated by how people have used NetBSD.
Linux can do that with User Mode Linux.
FYI rump is essentially unmaintained since its author left NetBSD years ago
Donated. I think what NetBSD is missing, but also hard to pull off, is something like https://openbsd.amsterdam. A service that people can use where parts of the fees go towards the foundation.
Thanks a lot :) agree with your idea.
Out of pure curiosity, is all the actual programming work for the foundation provided free of charge by volunteers? And the foundation expenses are mostly legal and administrative?
It wasn't clear to me based on the financials: https://www.netbsd.org/foundation/reports/financial/2023.htm...
Maybe the consulting section includes payments for programming work? Presumably at cheap rates, if so?
>Maybe the consulting section includes payments for programming work? Presumably at cheap rates, if so?
Have we reached the point on the timeline where we believe low-level operating system code should be acquired at "cheap rates"? While simultaneously, I assume, believing webshit cloud bollocks still demands top dollar?
I do not hold that belief, and nothing in my reply implies that I do.
Per the link I included, the total spent on consulting was 17,939.51 USD. So, if they were paying people, the people were working cheaply. But the consulting may have been for non-programming work. Hence my question.
Donated!! I run a small cluster of a few nodes I bought for cheap, and I’m experimenting with SSI on them. The kernel is really nice to read and modify.
Thanks a lot:)
Donated. I an thankful to NetBSD - I built some routers back in 2000. Long live NetBSD!
Thanks a lot :)
And let me guess... You haven't had to replace them since?
They don't exist anymore. No need to be sarcastic. But they gave me plenty of experience
Migrated them onto a toaster to reduce hardware footprint.
Donated! Thank you very much, NetBSD was one of my first experiences, on a Pentium 60 with a 504MB hard drive. It made me who I am today, eternally grateful to have learned from such amazing and talented people.
Thanks a lot :)
Do they offer a swag store like OpenBSD or FreeBSD? I realize they only get pennies from those sales but that’s typically my approach, buy a shirt for $30 and make an extra $20 donation.
Yes they do. Here is link https://www.freewear.org/NetBSD
Big fan of NetBSD. From the perspective of a kernel hacker, I found the NetBSD codebase to be very readable and easy to work with. FreeBSD and OpenBSD, while being more fully-featured and security-minded respectively, had to make compromises in their codebase as a result.
I just donated. It's important to keep projects like NetBSD vital, as a monoculture benefits nobody but the monoculturists. I think of it as a way to help ensure Survival of the Species - by diversity.
It's really in the best interest of everyone using it to chip in and keep the project relevant. Unfortunately the amount of donations is going to be contingent on the size of its user base which will need to grow to ensure its longevity.
eh, I don't use it, I chucked $150 their way.
Having NetBSD around is a net win, and the cost of doing business for them is extremely low for the product they provide.
Thanks a lot:)
an offspring to build your own minimal BSD UNIX system: https://www.smolbsd.org/
Does NetBSD really help reduce e-waste any more than Linux already does?
Maybe not yet but I can see Linux's place as the shitbox saviour start slipping a bit in the next few years. Debian dropping x86, distros getting fatter in general.. I can't really see those trends reversing. Meanwhile NetBSD goes against them.
However it goes, the main issue is one no operating system can solve which is modern life relying on the Web and beefier browsers. Unless you want to rebel against that you're probably better off getting a laptop from the past 10 years for < £100 on eBay.
Although I agree with beefier browsers, I also want to say that there are browsers like dillo etc. which can be good enough for simple websites and also not everything needs a web browser to be usable
Imagine this, a system which can watch movies, edit texts, create disks, have curl/wget, send and recieve files using piping server (which is a curl thing) , view pdfs, mpv and what not, a desktop manager, file manager etc.
As someone hacking around with the legendary tiny core linux, I am more and more mind blown each day with just how much can happen in 14-21 MB, you can definitely build a mini self hosting rack with just some remastering as tinycore can actually run podman as well (combine this with alpine containers to create a super duper minimalist self hosting things too)
the possibilities are endless. When I ran tiny core linux on my pc and ran nothing else, It took 21 mb in ram for a whole gui with editors and file managers etc. all running in ram so super fast filesystem with a package manager
I personally wanted to build my own operating system to limit myself to the most minimal system so taht I just study and do nothing else, I thought tiny core was it but then I tried to hack around it and there are sooooo many things in 21 mb, makes me appreciate minimalism
> which can watch movies
I have to say, the sheer fucking irony of this statement made me do a double-take.
I might be showing my age a bit, but I'm still remembering when web-browsing was considered a "light" activity (without extensions like Web Java), and watching a video was "very computationally expensive".
I guess some shift happened in the early 2010's where video playback was hardware accelerated more frequently; and complicated javascript started taking off as Google unveiled v8.
As much as the tech bro billionare class salivate over the idea of bringing an "everything app" to the market, we already have those. They're called browsers.
How is TLS negotiation and transport on older hardware (with no AES-NI hardware acceleration)?
I remember it used to be expensive as heck to do TLS back in 2014~, so much so that we bought accelerator cards and segmented "secure" servers so that load wouldn't hit the ordinary browsing of our sites...
"How is TLS negotiation and transport on older hardware (with no AES-NI hardware acceleration)?"
I use a TLS forward proxy. With today's overpowered hardware, I can even run the proxy on an old "phone" (but I cant run NetBSD^1)
This allows the older computers I own to use plaintext HTTP like the good old days
1. Despite https://en.wikipedia.org/wiki/T-Mobile_Sidekick
The argument for NetBSD is that it runs on almost anything that was ever produced. That isn't the case for Linux, even older x86 is no longer supported in the mainline.
That may be technically true but…
Linux (the kernel) may have been ported to more machines and architectures than NetBSD’s kernel, yes. But is all the code present in the same source tree or do you have to go find patch sets or unofficial branches?
More importantly: is there a modern distribution that builds an installable system for that platform?
The special thing about NetBSD is that you get the portability out of a single and modern tree for many more platforms than any single Linux distribution offers.
You said the same thing I did with extra steps.
That’s because… I misread your comment.
In any case, NetBSD is not well known and “why bother because Linux also runs everywhere too” so I thought it was worth explaining.
https://technically.kakwalab.ovh/posts/silly-sun-server-intr...
Some architectures are no longer practical with Linux. The kernel might still support it, but distribution support is sketchy.
For a SPARC64 server refurb project, the choices were pretty much OpenBSD or NetBSD in my case.
What do you think, there are milions of people or companies running NetBSD on 486 to protect the planet from e-waste? How many times have you replaced your phone with a newer model in the last 10 years?
yes. It supports like 60 different cpu architecture all back to 1979 VAX 790 INCLUDED.
Its also one of few OSes where 32-bit 386 is still tier 1 release.
All from single code source code tree.
Does running NetBSD on VAX from 1979 help to reduce e-waste, though?
Yes if you still run it. 0 metal wasted.
Some of the Linux distros are getting pretty fat and don't work so well on older hardware. Of course some are lean too. But NetBSD has a goal.
An interesting thread about the upgrade exists on the same list that the link points to, just a few messages earlier. Hopefully, with the donations, the project can improve the existing friction points - https://mail-index.netbsd.org/netbsd-users/2025/10/01/msg033...
Things like NetBSD seem like an obvious use case for tokens and DAOs (funding and governing treasuries for non-crypto open source software projects). Why is this not more common already?
OT what's with the email addresses with percent signs in them?
See http://www.faqs.org/docs/linux_network/x-087-2-mail.address.... (next to last paragraph)
“Then there is the % address operator: user %domainB@domainA is first sent to domainA, which expands the rightmost (in this case, the only) percent sign to an @ sign. The address is now user@domainB, and the mailer happily forwards your message to domainB, which delivers it to user. This type of address is sometimes referred to as “Ye Olde ARPAnet Kludge,” and its use is discouraged“
>Ye Olde ARPAnet Kludge
Seems fitting that NetBSD's internal mailing lists still use ossified address syntax from a time before DNS.
I would guess it's an anti-spam measure. Although if I'm reading sibling comment right, it is actually a valid email address? (Assuming you have a mail server running on localhost.)
It’s rather the opposite: Spammers used to exploit that mechanism back when it was more widely enabled.
This year I've seen some retro tech YouTube videos about people putting modern NetBSD in their expensive PDPs and Vax machines. Dave Plumber comes to mind.
Donated. While I don't use NetBSD, the existence of projects like this is essential for open source technology, operating system design, and the overall vitality of programming culture.
I'm curious what do people use NetBSD for?
It's my nameserver/DHCP server. I used to have it set up as an iSCSI target for backups and as a boot server for my firewall, but I do something else these days.
My main reason for using NetBSD for this is to have easy access to the man pages. Like the other BSDs, the man pages are exceptionally well-written and are a tremendous resource for doing POSIX programming. Plus I find myself digging through the code when I'm interested in how something is implemented. Having a local repository of good C code with a liberal license is worth having the extra OS to manage.
It was my desktop daily driver circa year 2000.
Not sure what people use it for now.
i ran NetBSD as a desktop on BeagleBoneBlack from circa 2010 until recently when I donated the hardware to a CompSci student.
...and as others mentioned it: the rump kernel.
For as long as I can remember, there was NetBSD and FreeBSD (OpenBSD and DragonFly came later).
I suppose after 30+ years, any chance of consolidation is hopeless and undesirable?
Code aside, the goals for each project are vastly different. There's nothing to be gained by consolidation.
Every Linux distro has different goals. But a unified kernel (more or less).
For hardware, can a single device driver be made for all variants of BSD? If so, then I agree.
You aren't going to see OpenBSD share a kernel with anyone - it's too different and makes trade-offs the others won't accept. And NetBSD doesn't need the heavyweight kernel FreeBSD uses.
From what I've seen, the BSD community swaps code around on a regular basis. But they pick and choose what code to use based on their own goals. It seems to work pretty well.
There's a lot of shared code.
What are the goals/specialization/actual use of each?
FreeBSD: I always got the impression this was trying to be full modern UNIX but non-linux
NetBSD: I guess this is for older/less powerful computers based on comments here?
OpenBSD: ???Security???
Dragonfly: a schism over threading, but FreeBSD?
That's so cheap compared to what they provide. Amazing.
Such an underfunded project. Even with such low resources they can get a lot done.
It surprised me that a project such as NetBSD only managed to raise ~$10k throughout the year. What's going on with the project?
The last few years or so, its activity seemed to have increased quite a bit, or maybe they are getting more press then they have had in many years :)
FWIW, this is the first time I have ever seen any mention of donations on any major tech WEB site.
What is NetBSD?
It's the amount of BSD you have from gross BSD after paying all the technical debts.
But then what does that make FreeBSD?
EBITDABSD.
It's a BSD variant dedicated to running on a wide variety of hardware.
One of the running jokes is that you can "run it on a toaster" — see https://news.ycombinator.com/item?id=45712368
If it can run DooM it can run NetBSD.
Not really. Doom just requires a 32 bit CPU (maybe someone even kluged it together on a 16 bit platform?) but NetBSD requires a CPU with an MMU.
NetBSD is the slim, small, traditional BSD that has an emphasis on clean code and portability. It's great for small jobs and it'll run on that old SPARC that's collecting dust in the closet. It's simpler than FreeBSD (the industrial strength BSD) and doesn't have the hyper focus on security that OpenBSD does.
Donated.
Thanks a lot:)
"I'm Doing My Part"
Thanks a lot :)
It’s so annoying that none of the corps using it aren’t putting a cent in and they ask individual developers to donate. Meh
That is the wonder of BSD like licenses for big corps.
If Linux never happened, we would still be using big iron UNIXes, each taking whatever they felt like from BSD variants.
Notice how all the new FOSS operating systems for IoT devices none of them use GPL, NutXX, FreeRTOS, Zephyr, Arduino libs, IDF,...
OTOH, if Linux never happened, much more work would be put into *BSD.
Absent something like OpenSolaris really taking off, the popular opinion in my circles is that *BSD would have (which is at least related to the same thing). Unless you believe Windows would have "won" which is certainly the side-bet a lot of companies were making at the time.
Not really, as anyone that was already using UNIX at the time is aware of, and the whole UNIX V6 vs BSD integrations as baseline for Aix, Solaris, Irix, HP-UX,....
No, we wouldn't. Linux climbed its way up to overtake proprietary UNIX despite being less capable, which it very much was at the time.
Linux came around at the right time when the Internet was going public and regular people had access to hardware that could run a decent UNIX. People latched onto it because it was free and an interesting project. The free BSDs were just late enough to the party that they missed out on the momentum.
All the proprietary UNIX vendors (other than SCO) relied on expensive proprietary hardware sales. Intel ate their lunch while they were too busy stabbing each other in the back to notice. Linux killed SCO because SCO was, quite frankly, overpriced crap.
None of this had anything to do with the license, other that the fact you could use it for free. It was all about hardware availability, the rise of the Internet, the wave of new IT people who had experienced Linux at home, and the fact that Linux on Intel was good enough to replace those pricy proprietary machines.
Now, you wanna talk Apple, there's where your code "theft" kicks in. But that's a whole different thing.
I was there, hence why it it easy to get quotes like these,
> 1998: Many major companies such as IBM, Compaq and Oracle announce their support for Linux.
From https://en.wikipedia.org/wiki/History_of_Linux
Without big money from UNIX vendors like those, cutting down their R&D costs, Linux would not have climbed anything.
GPL was the reason why they collaborated instead of being able to assimilate the code as they were doing with BSD, like anything sockets related.
Ironically IBM has recouped its investment, now as Red-Hat owners.
That is where everything on GNU/Linux that is mainly done by Red-Hat like GNOME, Gtk, GCC, Java is being paid for.
I wonder if there'll be some big cultural shift in open source as people get more annoyed at cases of big companies taking their code and giving nothing back/demanding them to work for free. Might already be happening just slowly
The people at the big companies who share the cause are not the ones who have anything important to decide there.
If you don't work for that big company, you're more than welcome to ignore any "demands" to work for free. And a lot of people in open source are indeed paid to work on it.
Maybe, but it seems to me that there's been a cultural shift away from GPL in recent years.
Which shows that the propaganda by bigtech companies over many, many years did work. :-(
AGPL and non-compete licenses like SSPL and FSL have been steadily adopted —Redis, ElasticSearch and Liquibase notably switched to them—, pretty much motivated by not wanting to be screwed over by Big Tech one way or another.
More than fine if you ask me, giving away your work for megacorps and oligarchs to steamroll your business or otherwise society at large isn’t much of a public service in the end
Yeah I have been thinking more and more about source available. I am not sure about AGPL as I feel like, some companies might still use it. honestly, it depends but all the companies which use non compete license somehow diverts back to AGPL if they were open source before
So if some company/product was open source and then used source available license, the backlash would be so much that they go to something like AGPl most of the time
but that happens because people feel betrayed because some might have contributed thinking its foss forever so its a rug pull
I think a good idea could be to have a source available license from the start so that everybody who ever contributes knows this as a fact.
What are your thoughts? What should I or anyone else pick? As a "foss" advocate, I would prefer AGPL but I don't want to get screwed by Big Tech ever with all the loopholes that they can have (like AWS), Honestly I don't know which is why I am asking really.
> What should I or anyone else pick?
My personal (non-mainstream) thought on this topic is: work on open source projects that serve a purpose that is very antithetical to the interests of bigtech companies. This way, such companies will be a lot less interested in "using" your project (without contributing back).
Honestly I feel like I've been seeing a lot more AGPLv3 for that very reason, and I'm 100% in support of it.
Are there any major corporations using NetBSD today?
I know in the past things like the network stack had been repurposed to other mainstream products.
Well NASA currently using pkgsrc of NetBSD for their NAS
https://www.nas.nasa.gov/hecc/support/kb/using-software-pack...
I doubt NetBSD gets much use by "big corps." It's used by hobbyists, researchers, and universities.
I'd hope at least some of the NetBSD developers are paid to work on the OS as part of their day jobs.
From Big Tech to startups, most corporations waste billions in ridiculous and frivolous projects and yet insignificant money is sent back to the projects they owe their very existence.
But don't you dare switch to a proprietary license or you will be dragged across social media as an evil selfish person. Even if it's only postponing source releases for a couple of years.
what is different than OpenBSD or just BSD???
"BSD" doesn't really exist as far as I'm aware. It was a proprietary operating system made at Berkeley for studying OS design.
OpenBSD/FreeBSD/NetBSD are open source continuations of the source-available 4.4-BSDlite code (removing AT&T proprietary extensions iirc).
OpenBSD follows BSD principles but focuses on code clarity and security.
FreeBSD tries to be very flexible, putting user-experience over security. (it has to be noted that OpenBSD is very usable, but lacks a lot of nice features like ZFS and DTrace that FreeBSD supports).
NetBSD is all about being incredibly lean and portable. NetBSD will run on basically anything, even things that Linux and other *BSD's have no hope of running on.
There absolutely was a real-world non-research operating system called just "BSD", and it ran primarily on DEC VAX hardware to start. Other Unixes like SunOS eventually started basing themselves on it. The last official Berkeley Software Distribution was 4.4BSD-Lite in 1994, which by then had been ported to many other platforms, including i386. The Computer Systems Research Group which maintained BSD disbanded the next year.
What about DragonFlyBSD?
DragonflyBSD is trying to be extremely progressive in their OS design by leaning in to how we're architecting new computers.
So, leaning in to how SSD's behave instead of how HDD's behave- ensuring that the kernel can make effective use of multiple cores etc;