flinux essentially had the architecture of WSL1, while CoLinux was more like WSL2 with a Linux kernel side-loaded.
Cygwin was technically the correct approach: native POSIX binaries on Windows rather than hacking in some foreign Linux plumbing. Since it was merely a lightweight DLL to link to (or a bunch of them), it also kept the cruft low without messing with ring 0.
However, it lacked the convenience of a CLI package manager back then, and I remember being hooked on CoLinux when I had to work on Windows.
Cygwin is way older than CoLinux. CoLinux is from 2004. Cygwin was first released in 1995.
The problem with Cygwin as I remember it was DLL hell. You'd have applications (such as a OpenSSH port for Windows) which would include their own cygwin1.dll and then you'd have issues with different versions of said DLL.
Cygwin had less overhead which mattered in a world of limited RAM and heavy, limited swapping (x86-32, limited I/O, PATA, ...).
Those constraints also meant native applications instead of Web 2.0 NodeJS and what not. Java specifically had a bad name, and back then not even a coherent UI toolkit.
Just use ssh from Cygwin. DLL hell was rarely a problem, just always install everything via setup.exe.
The single biggest problem it has is slow forking. I learned to write my scripts in pure bash as much as possible, or as a composition of streaming executables, and avoid executing an executable per line of input or similar.
As a dependency of a shipping Windows application that needs to cleanly coexist side-by-side with existing Cygwin installations and optionally support silent install/upgrade/uninstall through mechanisms like SCCM, Intune, and Group Policy?
Not so much.
I do use the setup program to build the self-contained Cygwin root that's ultimately bundled into my program's MSI package and installed as a subdirectory of its Program Files directory, however.
Slow forking is only the second biggest problem IMO. The biggest is the lack of proper signals. There's a bunch of software out there that just isn't architected to work well without non-cooperative preemption.
That's fake cooperative emulation of signals. It isn't preemptive (unless someone got a kernel driver approved while I wasn't looking?) thus many things either work poorly or not at all. Pause-the-world GC algorithms are a good example. Coroutine implementations also have to be cooperative.
If you're curious, I believe the issue was discussed at length in the Go GitHub issues years ago. Also on the mailing lists of many other languages.
I've never had a problem installing from setup, but some tools were (maybe still are, it is a long time since I've needed anything not in the main repo) ported to windows using the cygwin dlls were distributed with their own versions and could clobber the versions you have otherwise (and have their versions clobbered when you fix that).
> slow forking
There isn't much that can be done about that: starting up and tearing down a process on Windows is much more resource intensive operation than most other OSs because there is a lot going on by default that on other OSs a process ops into, only if it needs to, by interacting with GUI libraries and such. This is why threads were much more popular on Windows: while they are faster than forking on other OSs too, especially of course if data needs to be shared between the tasks because IPC is a lot more expensive than just sharing in-process memory, the difference is not as stark as seen under Windows so the potential difficulties of threaded development wasn't always worth the effort.
Cygwin can't do anything about the cost of forking processes, unfortunately.
Cygwin bash isn't slow either. The problem is a typical bash script isn't a series of bash operations, it's a series of command line program executions.
For example, someone might do something like this (completely ignoring the need to quote in the interests of illustrating the actual issue, forking):
for x in *; do
new_name=$(echo $x | sed 's/old/new/')
mv $x $new_name
done
Instead of something like this:
for x in *; do
echo $x
done | sed -r 's|(.*)old(.*)|mv \1old\2 \1new\2|' | grep '^mv ' | bash
This avoids a sed invocation per loop and eliminates self-renames, but it's harder to work with.
Of course the code as written is completely unusuable in the presence of spaces or other weird characters in filenames, do not use this.
To solve the problem or because you saw "slow" and "bash" and wanted to bring up something cool but unrelated?
If I go from 10 seconds of forking and .04 seconds of shell to 10 seconds of forking and .01 seconds of shell, I don't actually care about how cool and fast the shell is. And I've never had the speed of bash itself be a problem.
You don’t get an app that looks the same across platforms. You do get apps that look like they belong on your platform, even though the code is cross-platform. It uses the native toolkit no matter where you run it across Windows, GTK, Qt, Motif, macOS/Carbon, macOS/Cocoa, and X11 with generic widgets.
Older platforms are also supported, like OS/2, Irix, and OSF/1.
It’s a C++ project, but it has bindings for most of the languages you’d use to build an application. Ada? Go? Delphi? Ruby? Python? Rust? Yes, and more.
https://wiki.wxwidgets.org/Bindings
The problem is, most of these bindings are out-of-date. Delphi from 2012, Basic from 2002, D from 2016. wxRuby is a dead link. wxAda was already dead in 2009, as the discussion I can google suggests.
So, if you use wxWidgets, you probably have to use either C++ or Python version, others are unlikely to be supported.
I wish the Wx proponents would stop saying these things. Who exactly are you trying to fool? Do you have no concept of reputational damage? What good comes from a claim that is so easily disproven by just installing a Wx application and looking?
Do you understand the difference between a toolkit API and a graphical widget?
I’m not trying to fool anyone. I'm not affiliated with the project. I’m just aware of it and have used it a few times. You, on the other hand, have called me a liar and a fraud because I repeated exactly what the project docs state and which your two links do nothing to contradict. In fact, you linked to yourself being corrected by the actual maintainer of the project. Did you read anything he wrote?
> Do you understand the difference between a toolkit API and a graphical widget?
I think I do. I have taken a few minutes on the Web to compare that what I had in mind is correct. What was the point of asking this question? Was it to trap me in a gotcha, or paint me as clueless, or what?
> have called me a liar and a fraud because I repeated exactly what the project docs state
Good, you realise you are taking on the claims made by Wx on paper. However, there's more to the world. To get the full picture, you have to also engage with what I have listed. The docs say one thing, the reality shown in the screenshots say another. There is a contradiction. It remains unresolved, not for lack of trying on my part.
> your two links do nothing to contradict
You are not further allowed by me to invalidate what I was writing about by simply disregarding the evidence. Engage with the points I was making. The differences in look and feel between Wx and native are plain for everyone to see and verify. So, what now? Who is right?
> Did you read anything he wrote?
Yes. Examine this:
his claim> OTOH all the standard UI elements (buttons, checkboxes, text controls, date pickers, ...) are native
his deflection> Sorry, I don't know what is this supposed to prove
So instead of admitting that there is a contradiction, he just pretends to not understand it.
Also examine this:
> look good
> look good
> looks fine
> look good
I never mentioned anything about looking good, this is a distraction designed to deflect from the central point I was making. As I wrote before, the central point made by me remains completely unaddressed.
Alas, I cannot deal with those crazy-making techniques, his behaviour measured by outcome is indistinguishable from the mentally ill. With the help and advice from a friend, I came to the conclusion that it was not safe for me to respond, so I then decided not to.
I don’t really care what you “allow” me to do, and I don’t owe you my time for a deep rebuttal of your claims.
I accept that you’ve had a bad experience with the project and that you are speaking in good faith about that experience. I do not accept that I’m a liar or a shill because my experience does not match yours, and you will not convince me to accept that.
Whatever has gone on in other places between you and some third party has nothing to do with me. I didn’t see any credible threat to your safety in the links you shared, but I’m willing to guess that’s not the complete story between the two of you. I really can’t and won’t be dragged into the middle of that, because I don’t know either of you.
I notice you did not address the central point. This is, of course, because you are unable to. You have no argument to attack the evidence I have shown, you cannot answer any of the questions I have posed, the sidetrack about the other HN user was fruitless, you rationally know you are holding on to an indefensible position. And yet… the fragile ego cannot allow admission of being wrong, or even merely entertain the thought. So "muh experience is different" emotional cope comes out instead of dealing with objective reality, like installing a Wx software or looking at a screenshot. Stop digging, man, we already feel embarrassed on your behalf.
If you can read someone’s thoughts so thoroughly you should be making a living playing poker. All you have cited as evidence is your own other posts. You just have to have the last word, though, and that word has to be an insult to a stranger.
I used cygwin pretty heavily in the late 90s and early 2000s. It was slow. I had scripts that took days to run dealing with some network file management. When I moved them over to linux out of frustration (I think I brought in something like a pentium 90 laptop, gateway solo I think?) they were done in tens of minutes.
I'm sure they did the best they could ... it was just really painful to use.
This matches me experience as well. Some of my earliest rsync experiences were with the Cygwin version and I can remember scratching my head and wondering why people raved about this tool that ran so slowly. Imagine my surprise when I tried it on Linux. Night and day!
It's not just DLL hell. Cygwin was also notorious for being really out of date. Security vulnerabilities and missing features were both very common at one point.
I have used cygwin for 30 years and never had any dll hell issues, because all the programs came from the cygwin installer. Never once needed something outside it.
Meanwhile those that complained about Java, now ship a whole browser with their "native" application, and then complain about Google taking over the Web.
Technically correct by some estimation, perhaps, but Cygwin is a crazy approach, was slow (contrary to the implication of the "low cruft" claim), was not as compatible as these other approaches, required recompilation, and was widely disliked at most points in its life. There's a lot of crazy voodoo stuff happening in cygwin1.dll to make this work; it totally qualifies as "hacking in some foreign Linux plumbing", it's just happening inside your process. Just picture how fork() is implemented inside cygwin1.dll without any system support.
Cygwin doesn't work at all in Windows AppContainer package isolation; too many voodoo hacks. MSYS2 uses it to this day, and as a result you can't run any MSYS2 binaries in an AppContainer. Had to take a completely different route for Claude Code sandboxing because of this: Claude Code wants Git for Windows, and Git for Windows distributes MSYS2-built binaries of bash.exe and friends. Truly native Windows builds don't do all the unusual compatibility hacks that cygwin1.dll requires; I found non-MSYS2-built binaries of the same programs all ran fine in AppContainer.
A lot of this is issues Microsoft could fix if they were sufficiently motivated
e.g. Windows lacks a fork() API so cygwin has to emulate it with all these hacks
Well, technically the NT API does have the equivalent of fork, but the Win32 layer (CSRSS.EXE) gets fatally confused by it. Which again is something Microsoft could potentially fix, but I don’t believe it has ever been a priority for them
Similarly, Windows lacks exec(), as in replace the current process with new executable. Windows only supports creating a brand new process, which means a brand new PID. So Cygwin hacks it by keeping its own PID numbers; exec() changes your Windows PID but not your Cygwin PID. Again, something Microsoft arguably could fix if they were motivated
> A lot of this is issues Microsoft could fix if they were sufficiently motivated...
They did fix it, in a sense, with WSL1 picoprocesses. Faster and more compatible than Cygwin. Real fork and exec on the Windows NT kernel. Sadly, WSL2 is even faster and more compatible while being much less interesting. WSL1 was pretty neat, at least, and is still available.
In any event, this diversion doesn't change my analysis of Cygwin. Cygwin still sucks regardless of whose fault it is. I intentionally left this stuff out of my post because I thought it was obvious that Cygwin is working around Windows limitations to hack in POSIX semantics; it's the whole point of the project. None of us can change Windows or Cygwin and they're both ossified from age and lack of attention. We have to live with the options we've actually got.
If you need a Windows build of a Linux tool in 2026 and can't use WSL, try just building it natively (UCRT64, CLANG64, MSVC, your choice) without a compatibility layer. Lots of tools from the Linux ecosystem actually have Windows source compatibility today. Things were different in the 90s when Cygwin was created.
Developing on cygwin, however, was a right pain. If a C library you wanted to use didn't have a pre-built cygwin version (understandable!) then you end up doing 'configure, make' on everything in the dependency tree, and from memory about two thirds of the time you had to edit something because it's not quite POSIX enough sometimes.
Ha ha doing Unix like it was 1989. At the time I thought configure was the greatest of human achievements since I was distributing software amongst Sun machines of varying vintage and a Pyramid. I want to say good times but I prefer now ha ha
Autotools was designed to produce a configure script with zero dependencies other than the compiler toolchain itself. I always thought it would be a good way to bootstrap a system configuration database (like the kind X11 already had, the name I forget) but it turned out to be too convenient to just drop autotools into every project instead.
So now even today, compiling any GNU package means probing every last feature from scratch and spitting out obscenely rococo scripts and Makefiles tens of thousands of lines long. We can do better, and have, but damn are there a lot of active codebases out there that still haven't caught up.
Reminds me of a fun weekend I spent ~5 years ago building the newest version of every GNU program I could get to build on NEXTSTEP 3.3 (running on 68k NeXT hardware) without major changes.
Nowadays MSYS2, which does depend on cygwin under the hood, offers such a package manager (pacman of Arch Linux) and it is quite a user friendly to run native POSIX binaries on Windows without a linux VM.
In my personal experience, Msys 2 would work great until it didn't. Unless this has changed, from what I remember, Msys2 compiled everything without PIC/PIE, and Windows does allow you to configure, system-wide, whether ASLR is used, and whether it's used "if supported" or always. If that setting is set to anything but off, Msys2 binaries will randomly crash with heap allocation errors, or they do on my system. It happened so much to me when I had actual coreutils installed that I switched to uutils-coreutils even though I knew that uutils-coreutils has some discrepancies/issues. Idk if they've fixed that bug or not; I did ask them once why they didn't just allow full ASLR and get on with things and they claimed that they needed to do non-ASLR compilations for docker.
MSYS2 is very confusing. When you pick "MSYS2", you are building exclusively for the MSYS2 target environment, and might not have proper compatible windows headers. When you pick "MINGW32/64", you are instead building for the normal windows environment, and get proper windows headers. But if you didn't know that, you would end up confused about why your program is not building.
It doesn't help that the package simply named "gcc" is for the MSYS2 target.
And just to add insult to injury, you probably don't want MINGW64 either, as it relies on the ancient MSVCRT.DLL C runtime library that lacks support for "new" features like C99 compatibility and the UTF-8 locale, and that Microsoft never supported for use by third-party applications in the first place.
Instead, you either want UCRT64 or CLANG64, depending on whether you want to build with the GNU or LLVM toolchains, as it uses the newer, fully-supported Universal C Runtime instead.
It's still useful to use MSVCRT in certain circumstances, such as targeting the earliest 64-bit versions of Windows.
As for UTF-8 support, it's the manifest file that determines whether Windows sets the ANSI code page to UTF-8. (There's also an undocumented API function that resets the code page for GetACP and the Rtl functions that convert ANSI into Unicode. But this would run after all the other DLLs have finished loading.) Having the code page correct is enough to support Unicode filenames and Unicode text in the GUI.
It just won't provide UTF-8 locale support for the standard C library.
Sure, or older 32-bit versions of Windows for that matter, or for building software that hasn't been ported to UCRT.
I can certainly relate to this: I'm currently sitting on a request for an enhancement to a product (currently running on a 32-bit Windows 10 VM) with a build system that has never been updated to support any Microsoft platform other than MS-DOS, or toolchain newer than Microsoft C 5.1.
> lacks support for "new" features like C99 compatibility
This made me laugh. It reminded me of course work I did in university that was clearly written many years before I took the course as it recommended we manually enable the "new" c99 standard in our compiler, which I guess survived in the documentation up through when I took the course, at which point it was still relevant since GCC was otherwise defaulting to C11 by the time I was using it.
It's annoying to wade through six different versions of the same package for different runtimes and word sizes. Heaven forbid you accidentally install the wrong one.
Cygwin implements a POSIX API on Win32 with a smattering of Nt* calls to improve compatibility but there's a lot of hoop jumping and hackery to get the right semantics. Fork isn't copy on write, for one thing.
I was a Cygwin user from about 1999 to 2022 or so, spent a little time on wsl2 (and it's what I still use on my laptop) but I'm fully Linux on the desktop since last year.
Ha that tracks my own usage and timeline almost precisely, although I was using cygwin and WSL2 in parallel for a while. Lot of complaints about cygwin speed here, but NTFS filesystem access is actually a lot faster on cygwin than WSL2!
It's kind of both. Hyper-V is a bare-metal (type 1) hypervisor. Windows runs virtualized, one level above it, in a privileged (host) VM, next to other (guest) VMs.
Requiring every single Linux app developer to recompile their app using Cygwin and account for quirks that it may have is not the correct approach. Having Microsoft handle all of the compatibility concerns scales much better.
Why not? That is just a matter of porting stuff over, like a FreeBSD ports collection, an apt repo, or a bunch of scripts for Proton/Wine such as Lutris.
Cygwin started in 1995. Microsoft wasn't cooperative with FOSS at all at that point. They were practicing EEE, and eating some expensive Unix/VMS machines with WNT.
Assuming you were on NT-lineage, rebuilding for SFU (Interix) was the technically correct and nice implementation, though since a lot of Linux programs are non-portable (or have maintainers who mistakenly think they can do better than autotools) it was a pain in practice.
So, is it like colinux[0], but for pre-NT windows? Neat!
Back when I was still using windows (probably XP era), I used to run colinux, it was kind of amazing, setting up something like LAMP stack on the linux side was a lot easier and then using windows editors for editing made for quite nice local dev env, I think! Could even try some of the X11 servers on windows and use a linux desktop on top of windows.
When I noticed I kept inching towards more and more unixy enviornment on the windows, I eventually switched to macOS.
Apart from the obvious hack-value, I can't quite imagine even pretend use-case, with some 486 era machine, you would be limited by memory quite quickly!
What I especially like about this Windows 9x Subsystem project, is that it proves that coLinux could have been written way earlier. Now imagine how less dual booting we would have needed in 1996 if that happened, and how it would have affected VMware which only existed since 1998.
I remember myself on my first year of CS, set theory classes, at the whiteboard, trying to make a proof, but there something I was not able to prove at all, so I said 'it's trivial' and the doctor said 'yeah, it's trivial' and we went further.
The one I've always flown with is, trivial means (1) a special case of a more general theory (2) which flattens many of the extra frills and considerations of the general theory and (3) is intuitively clear ("easy") to appreciate and compute.
From this perspective, everything is trivial from the relative perspective of a god. I know of no absolute definition of trivial.
No, this was really something trivial, in the sense that you could feel it's true. Like 2+2=4 but to prove it you need to create a set of functions, axiom and a theorem
> But I wonder how it seems to people who understand how it works?
As someone who mostly understands what's going on - It does not seem like wizardry to me, but I am very impressed that the author figured out the long list of arcane details needed to make it work.
The primary function of modern operating systems is to allow multiple programs to run, without interfering with each other, even if they try too. This means that each program can only read its own limited amount of memory and only gets to use the processor for a limited time, before another program gets a turn. Windows did not start using those features until Windows NT, which XP is based off of. Through Windows 98, any program could do whatever it wanted, and that hardware sat idle. Windows versions up to 98 were more like a library of features that a program could use, to display a user interface and talk to hardware peripherals.
There's special hardware in a processor, for the operating system to limit each programs access to memory and processing time, which Windows 9x leaves unused. This means that the Windows 9x Subsystem for Linux can say "look at me i'm the operating system now" and take over that hardware to run a modern operating system.
Windows 3.11 was a hypervisor running virtual machines. The 16-bit Windows virtual machine (within which everything was cooperatively multitasking), the 32-bit headless VM that ran 32-bit drivers, and any number of V86 DOS virtual machines.
Win9x was similar in the sense that it had the Windows virtual machine running 32-bit and 16-bit Windows software along with V86 DOS VMs. It did some bananas things by having KERNEL, USER, and GDI "thunk" between the environments to not just let 16-bit programs run but let them continue interacting with 32-bit programs. So no, Win9x was in fact 32-bit protected mode with pre-emptive multitasking.
What Win9x prioritized was compatibility. That meant it supported old 16-bit drivers and DOS TSRs among other things. It also did not have any of the modern notions of security or protection. Any program could read any other program's memory or inject code into it. As you might expect a combination of awful DOS drivers and constant 3rd party code injection was not a recipe for stability even absent bad intentions or incompetence.
Windows 2000/XP went further and degraded the original Windows NT design by pulling stuff into kernel mode for performance. GDI and the Window Manager were all kernel mode - see the many many security vulnerabilities resulting from that.
This is correct. Win9x did have memory protection, it just made an intentional choice to set up wide open mappings for compatibility reasons.
WSL9x uses the same Win9x memory protection APIs to set up the mappings for Linux processes, and the memory protection in this context is solid. The difference is simply that there is no need to subvert it for compatibility.
That's greatly oversimplified, or less generously, just flat out wrong. Win32 programs have always had their own isolated address space. That infamous BSOD is the result of memory protection hardware catching an access to something outside of that address space. When you open a DOS box, it uses the paging and V86 hardware mechanisms to create a new virtual machine, even though it shares some memory with the instance of DOS from which Windows was booted.
What Windows 9x didn't have was security. A program could interfere with these mechanisms, but usually only if it was designed to do that, not as a result of a random bug (if the entire machine crashed, it was usually because of a buggy driver).
It's mostly explained if you go to the project page. For me, the I would say the hardest thing about something like this is gleaning the Microsoft driver APIs. In the 9x days, Microsoft documentation was not quite thorough and difficult to access. It's still not pleasant.
This is in the class of things where even if the specific text doesn't trace to a true story, it has certainly happened somewhere, many times over.
In the math space it's not even quite as silly as it sounds. Something can be both "obvious" and "true", but it can take some substantial analysis to make sure the obvious thing is true by hitting it with the corner cases and possibly exceptions. There is a long history of obvious-yet-false statements. It's also completely sensible for something to be trivially true, yet be worth some substantial analysis to be sure that it really is true, because there's also a history of trivial-yet-false statements.
I could analogize it in our space to "code so simple it is obviously bug free" [1]... even code that is so simple that it is obviously bug free could still stand to be analyzed for bugs. If it stands up to that analysis, it is still "so simple it is obviously bug free"... but that doesn't mean you couldn't spend hours carefully verifying that, especially if you were deeply dependent on it for some reason.
Heck I've got a non-trivial number of unit tests that arguably fit that classification, making sure that the code that is so simple it is bug free really is... because it's distressing how many times I've discovered I was wrong about that.
[1]: In reference to Tony Hoare's "There are two ways to write code: write code so simple there are obviously no bugs in it, or write code so complex that there are no obvious bugs in it."
I don't think you understand how jokes work. They are mostly "distortions" of real dialog or events to add incongruous or absurdist elements. Here, Hardy's not uncommon momentary doubt about whether a statement really was obvious, while faintly amusing, is made into a joke by turning the momentary doubt into a 15 minute excursion. People then riff on the joke by turning that excursion into a mathematician presenting an elaborate proof that a statement is "obvious", quite contrary to the meaning of "obvious".
This being on the front page the same day as "Show HN submissions tripled and now mostly have the same vibe-coded look" is nice to see. One person spends 6 years understanding Win9x internals to run a modern Linux kernel inside it. The other thread is full of apps that took 20 minutes to prompt into existence. Posts like this make me happy.
(not trying to pick a fight, but...) what difference it makes if an agent spent 100$ to understand and came to the same result, given it is those who use it that will benefit of it, and thus the craze here.
note: i'm not saying author did not improve his skills overall, but also last '6 years' perhaps also means - fair amount of digging the web with search engines, which are... like AI 0.1
The difference is that despite the fact that I will never use this, I am nonetheless here to celebrate the effort. What difference does it make if someone runs a marathon or takes an uber given they may arrive at the same location?
Running marathon is not the same as getting to a predestined destination. Try running Rome-Paris next time instead of taking the train, I guess is the same. But still not the marathon goal.
And likely the prompt itself was generated. Instead of "create me an owl app" write: "create me a comprehensive prompt to create an owl app". Then paste that prompt to next AI session.
I've been working on a WebExtension that calls out to zygolophodon and returns plain HTML to the browser. In the process of rebasing it over recent changes but here is the working webext-old branch:
If I can get this to work (haven't tried yet) it directly solves a problem I have right now this week right here in 2026, 30 years after Windows 95 was even a thing.
Yes, I have weird problems. I get to look after some very weird shit.
If you're dealing with weird legacy 9x systems in 2026, another headache you've probably run into is getting them to talk to the modern web (since modern TLS and JS completely break old browsers).
I actually built a win9x compatibility mode into BrowserBox specifically for this kind of weirdness. You run the server on a modern system and launch it with bbx win9x-run, and it proxies the modern web to legacy clients. It works surprisingly well with IE5, IE6, and old Netscape on Windows 95/98/NT. Might be a fun addition to your retro utility belt!
Kernel-Ex and Basilisk and friends (Serpent browser or whatever it's called) plus some TLS stuff can browse the modern web just fine. JS, OTOH... get NoScript ASAP and block selectively.
Or better, ditch the web completely and head to Gopher/Gemini.
I agree there seem to be other options. But they all have fractal-edged interfaces to the legacy stuff, and you need to be careful about, as you said, script issues and vulns. That's the BrowserBox advantage - a fully modern, remote rendering system that can still be accessed from the legacy box, but securely. And easy setup. Admittedly, doing it the way you're doing is probably free - BrowserBox is not free - but it saves you lots of this delicate management of config, etc. Because the interface model is cleaner: just one HTTP endpoint that legacy browsers can actually access, streaming them fully modern browsing that's rendered elsewhere, securely.
I get the degradation to Gopher as a way to solve many issues, but many things just don't work there. And it may have its own vulnerabilities.
Ah! I solved that easily enough. Internet Explorer 8 works okay for the webby front ends to operate the transmitter equipment. How do you run that, in this day and age, safely?
Run it in Windows XP, in a VM.
Now here's the clever bit - qemu will allow you to expose the keyboard, mouse, and framebuffer as a VNC server. So you set up Apache Guacamole to point a VNC client at the VM, and then "normal people" can log in, operate the transmitter, and log out again.
You can do a lot of sneaky things with that, including setting up headless X, running VNC on it pointed at your qemu VM, and then streaming the headless X servers's framebuffer out with ffmpeg.
Yes sometimes work can be a bit boring with not much to do, why do you ask?
That is pretty cool, man. VM with XP in QEMU sending VNC frames to Guacamole clients on the web.
BrowserBox is basically the same pattern as this setup (streaming graphics from some browsing substrate somewhere to web clients) except architecture is different: a modern box on the same private network runs the BrowserBox server, and the win box (QEMUd or otherwise) connects to its http endpoint, using whatever browser it has (tested back to IE5 even, tho that's a way more buggy browser than IE6. IE8 should be golden). That way you get the full modern web, no compromises. But crucially the web is not actually accessing your legacy box. So, no comrp0mises, ie., no easy vulns. Especially a concern for older browsers. Plus, we've got policy controls to lock down capabilities (copy, paste, URL lists, internal IP access controls, etc).
In your case it sounds like you are running the webby servers on the XP box, too, so BrowserBox would link back into those over the same private network, render them on the modern box, send it back to the XP box, then clients can connect over the QEMU VNC bridge you already have.
Alternately you could just do away with the Win XP browsing, and have BrowserBox connect to your webby endpoints for transmitters wherever you run them, and then expose that browsing graphics stream to clients over whatever endpoint you want. Many options!
I like your ffmpeg out setup. How did that go? Share more about that? Pretty interesting, I love this old architectures, and legacy systems compatibility quests.
I should do a blog post and stick it on here, right?
> In your case it sounds like you are running the webby servers on the XP box, too,
Yeah - it's a web front end to some specialised software written on I guess Microsoft C++ (if I had time, enthusiasm, and a copy of it lying around I suppose I'd wave Ghidra at it and see what happens).
I'll look into BrowserBox, that sounds handy.
> I like your ffmpeg out setup. How did that go? Share more about that? Pretty interesting, I love this old architectures, and legacy systems compatibility quests.
Surprisingly well. I think you could probably stream it straight to Twitch or something if you wanted. Yeah, this sounds like a blog post.
There is a section of the Forties Pipeline where they have a huge amount of gas handling plant in central Scotland. Last time I was on site (admittedly 15 years ago but I don't see this changing soon) the SCADA outstations were run by absolutely minty box fresh VAXStation 3100s. Plastic not even peeled off the front panel badges fresh.
I recently saw that running on special 16 channel DAT recorders used by the 999 service, recently as in "within the past five years". I believe they've been retired but kept around in case they need to recover tapes off them.
I kept my mouth ABSOLUTELY FUCKING SHUT about knowing my way round OS/2 Warp 4.
I've completely forgotten pretty much everything about OS/2... worked support for iomega for a while covering OS/2, Jazz and 2nd level zip calls in the late 90's. I have very little desire to look back to those days.
Except it would be nice if computers were as snappy as back then... so many layers of cruft that nothing really pops. I remember I used to just disable all the NT/2K animations and it felt so nice. I built a career on web apps, but most are just really poorly written it pisses me off.
What's a good resource going over the architecture of Windows 3.x and 9x? I know bits and pieces, like that it has a "VM Monitor", and there's support for this sort of thing, though the details are all over the place. Most people summarized Windows as just "running on top of DOS", which is clearly not correct. Obviously, it doesn't use "virtual machines" in exactly the modern sense of the word, but there's clearly something cool and technical going on, that most sources seem to gloss over.
Before Windows Internals book we all know (called Inside Windows NT in its first editions), there was Windows Internals and other books and articles by Matt Pietrek. It starts from disassembly of WIN.COM studying the insides of DOS to figure out under which more or less common version of MS-DOS Windows is being run.
By Microsoft’s convention, that would be a way to run Windows 9x on Linux. It’s a bit confusing. Another example is “Windows Subsystem for Android”, which is what they use for running Android apps on Windows. I think the idea is that it’s not a “Windows Subsystem” for X, but rather a Windows “Subsystem for X”.
Which is exactly what the post says this is. It's running Windows 9x on Linux kernel. It's strangely worded, but from the follow up comment, and the readme in the repo it seems clear that it's running on the Linux kernel.
No it's the Word app which is included in your organization's Copilot 365 .NET SharePoint license assuming you know how to trigger the download. The chatbot the license is named after doesn't have a clue either so good luck.
It runs Linux with Windows underneath it, hence Windows is the subsystem being subordinate (in the most literal sense where it simply means "order" with no further implications) to Linux.
Per wongarsu's post, something like the OS/2 Subsystem is an OS/2 system with Windows beneath it, but the OS/2 Subsystem is much smaller and less consequential, thus subsidiary (in the auxiliary sense) to Windows as a whole.
Isn't marketing fun?
This is how we end up with hundreds of products that provide "solutions" to your business problems and "retain customers" and upwards of a dozen other similar phrases they all slather on their frontpages, even though one is a distributed database, one is a metrics analysis system, one handles usage-based billing, one is a consulting service, one is a hosted provider for authentication... so frustrating trying to figure out just what a product is sometimes with naming conventions that make "Windows Subsystem for Linux" look like a paragon of clarity. At least "Linux" was directly referenced and it wasn't Windows Subsystem for Alternate Binary Formats or something.
I agree. Don't have a citation now, but I remember reading that this was a copyright problem. They wanted to name it "Linux Subsystem for Windows", but apparently the Linux foundation does not allow unaffiliated projects to have a name beginning with "Linux", or something like that.
doslinux is some tricky sleight of hand where it looks like Linux is running inside DOS, but it's actually the other way around (even though DOS boots first).
WSL9x takes quite a different approach. Windows boots first, but once Linux starts both kernels are running side-by-side in ring 0 with full privileges. They are supposed to cooperate, but if either crashes then both go down.
Everytime I see something like this, I'm like, how in the hell did they learn and then figure this out? Congrats on this!!!! I will definitely have to play with this for some of that sweet nostalga.
I know there’s a small hobby industry of people creating Windows 9x compatibility shims for later Windows APIs… but who needs that when you can just run Linux apps?
Little late but would this have actually allowed running early Linux under Windows when Windows 95 came out in the 90s? I remember only dual booting being available at that time.
Hmm I wonder how stable it is.. It cannot render correctly Window control buttons (Minimize, Maximize, Close). If it fails on such basic task, I wonder where it crashes...
I remember running CoLinux back in my training days ~2005-2006. It seemed a bit like black magic to me at the time, but it worked quite well for my needs.
But CoLinux - IIRC - required the NT branch of Windows. I can only imagine the level of hackery it takes to make this happen on Windows 9x.
Part of me wants to weep at the sheer perversity, part of me wants to burst into manic laughter. It is indeed a world of endless wonders.
Interesting idea. Does it actually work? If so, didn't the guy kind of simplified WSL here?
I have not tested this yet, so I have no idea - but if he managed to pull this off then this may be one of the greatest achievements this year. Or perhaps there are some restrictions to it? Does compiling stuff work in it? So many questions ... who has the answers?
The core of the software is a subsystem, specifically a Windows subsystem; you're not running this subsystem on macOS or FreeBSD.
The "for Linux" is added because it's a subsystem for Linux applications (originally not leveraging a VM).
Microsoft also had the "Microsoft POSIX subsystem" (1993) and "Windows Services for UNIX" (1999) which were built on the "Subsystem for Unix-based Applications" (rather than "Unix-based Application Subsystem"). That chain of subsystems died at the end of Windows 8, though.
There are many reasons not to put "Linux" in front, but the naming is consistent with Microsoft's naming inconsistencies. It's not the first time they used "subsystem for" and it's not the first time they used "Windows x for y" either.
The naming is ambiguous, you could interpret the Windows subsystem for Linux as a subsystem of Linux (if it had such a thing) that runs Windows, or as a Windows subsystem for use with Linux. Swapping the order doesn't change that.
In other languages, the difference would be clearer.
No natural language is inherently imprecise. Every language has its own system to resolve vagueness or ambiguity and elaborate on the supposedly "missing" features of the language. This issue is relatively settled in linguistics.
My personal experience with our legal department on naming is that if your product name includes someone else's trademark, you have to say "Our Thing for Their Thing", exactly like that. I was involved in a product that did this, and we came up with some better names, but legal said no, it must be named with "for Their Thing" at the end. Those were the magic words so we don't get sued, and indeed, we weren't sued. Our legal was non-technical and never heard of WSL; they came to this conclusion independently.
The name we shipped was even worse than Windows Subsystem for Linux, honestly. At least Microsoft spent some time on it.
I always have the same problem myself. Same as I had with version naming of old programs like "Microsoft Word for Windows 2.0" instead of the easier "Microsoft Word 2.0 for Windows".
The problem is Word 2.0 for Dos was released in 1985, and Word for Windows 2.0 was released in 1991. Calling it Word 2.0 for Windows wouldn't make sense, because it wasn't the 1985 release with a new coat of paint, Word for Windows was its own thing, and this was the second version. Word for Mac was also separate, but eventually Word 6 was a common code base and it made sense to have Word 6.0 for [whatever]
Other people already answered but windows was just another personality on the original idea that cutler had for WNT. It just took a while for it to get implemented as a linux
To reciprocate the naming of Wine, maybe it could have been named Line. Also, both have this positive clang, being associated with "having a good time".
WSL was a subsystem to run Linux under Windows 10 without an explicit VM host. It was an Ubuntu Linux version. Later MS released WSL2, complete redesign, still for Windows 10.
With Windows 11 the WSL2 also supported GUI programs.
The NT kernel (used in NT 3.1 through 2000, XP, and eventually backported to Windows 10/11's WSL) was designed from scratch in 1993 with a POSIX subsystem from day one. The whole design philosophy was "multiple personalities" - handle syscalls from different environments by translating them to native NT kernel calls. WSL1 in 2016 essentially reimplemented that same trick for Linux.
Windows 9x, by contrast, was DOS-derived. Running Linux inside it would require fundamentally different (and messier) hacks - which is probably why nobody did it at the time. The very fact that this works at all is a testament to how ahead-of-its-time NT's architecture was.
For a practical answer: you'd need something like this for legacy locked-in situations. Old medical or industrial software that only runs on Windows 98, or specialized hardware without modern drivers. That said, if you have a 486 handy in 2026, running Linux natively is almost certainly more useful than running it inside a 30-year-old DOS derivative.
Oddly enough, I could kind of use this right now. I have some software which used SCSI (Adaptec WNASPI32.dll) calls to administer a device over the SCSI bus .. would this Subsystem be usable for that, or does it still require I build a WNASP32.dll shim to do translation?
So, you have Windows software. This "Windows 9x Subsystem for Linux" just boots Windows 95. I don't know what you would use the Linux part for. Care to explain more what you want to do?
If you want to run your windows software in Linux, you could try Wine. Wine seems to have support for WNASPI so it's possible your software would just work. (You might have to run Wine as root I guess, to get access to the SCSI devices.)
If Wine doesn't work, Windows in QEMU with PCI passthrough to the SCSI controller might have better chances to work.
I need raw SCSI I/O to be virtualizable in the linux context, so I could run a Windows app (yes it already works in Wine), and have it 'see' a SCSI device as if it were real hardware.
Wines WNASPI32.dll is really just a facade - it doesn't provide actual SCSI services, its just there for SCSI-using apps to think they have ASPI onboard - so for my case I would need to write a shim to pass through SCSI IO requests to a Linux service - or loopback file? - to actually process the requests. I've been meaning to do this for a long time, but if there is some way I can set up a loopback file under Linux to 'pretend' to be a SCSI block device for a Windows app, I'd sure like to know if its possible ..
This. The post immediately reminded me of Win4Lin 9x (the version before it became just another boring VM) and SCO Merge. It was insanely fast, even on the hardware of the day.
The Wikipedia page is not verify informative and presents it as a regular VM (possibly mixing up 9x and later versions that run the NT line of kernels). The manual is a bit more informative about the tech:
I’m a bit surprised it hasn’t been mentioned a lot in the comments. Maybe it’s a bit too old for most people here (Linux in the late 90ies/early 00s was a much smaller community)?
I think this is in the sense of needing a modern-ish hardware with VT-x/AMD-V support (instead of the already-contemporary v8086 which is already in use by Windows at this time).
lol, if you do assemble that Pentium desktop, one of the first things you'll notice is that the modern web is completely unusable on it natively.
To get around that, I recently added a legacy compatibility mode to BrowserBox (bbx win9x-run). It basically lets you run the server on your modern daily driver, and access it via IE 5, IE 6, or Netscape on the Pentium box. It strips away the modern TLS/JS rendering issues and lets you actually browse the modern web from Windows 9x. Highly recommend giving it a spin if you get that machine built!
Sidebar: does anyone else find the naming of this super confusing—does this mean embedding windows in linux or linux in windows? Surely we could find a way to refer to composing operating systems that isn't inherently ambiguous. Say, "linux inside of windows" would be less ambiguous.
Edit: to the people who downvoted, I corrected my spelling mistake. I will perform due penance.
Wait until you find IE was released for Unix, using some Win32 shims. And... die hard Unix sysadmin ran it under FVWM and compared to Netscape wasn't half bad.
Both propietary, but sadly NScape didn't open Mozilla yet, and the rest of the alternatives such as Arena/Amaya coudn't compete with 'modern' CSS features and the like.
Speaking of vintage IE and Netscape on old Win, it's actually still possible to use them to browse the modern web if you proxy it.
I built a Win9x compatibility mode for BrowserBox that does exactly this (https://github.com/BrowserBox/BrowserBox/blob/main/readme-fi...). Ur modern server does all the rendering, and it outputs a client link specifically designed for legacy browsers like IE5, IE6, and Netscape running on Windows 95/98/NT, streaming them the pixels. It's definitely an abomination, but there's something magical and retro that I like about viewing the 2026 internet through an IE6 window ;) ;p xx
- Retrozilla with some about:config flags disabling old SSL cyphers and new keys to enable
newer ones
- Iron TCL maybe with KernelEx and BFGXP from https://luxferre.top reading gopher and gemini sites
such as gemini://gemi.dev proxying all the web bloat and slimming it down like crazy
Not sure if I’d call converting modern web to Gemini better. It’s a downgrade. If you want modern fidelity on old systems you need BrowserBox. Or i guess you could fiddle with Retrozilla flags if that works
http://www.colinux.org/
https://github.com/wishstudio/flinux
flinux essentially had the architecture of WSL1, while CoLinux was more like WSL2 with a Linux kernel side-loaded.
Cygwin was technically the correct approach: native POSIX binaries on Windows rather than hacking in some foreign Linux plumbing. Since it was merely a lightweight DLL to link to (or a bunch of them), it also kept the cruft low without messing with ring 0.
However, it lacked the convenience of a CLI package manager back then, and I remember being hooked on CoLinux when I had to work on Windows.
The problem with Cygwin as I remember it was DLL hell. You'd have applications (such as a OpenSSH port for Windows) which would include their own cygwin1.dll and then you'd have issues with different versions of said DLL.
Cygwin had less overhead which mattered in a world of limited RAM and heavy, limited swapping (x86-32, limited I/O, PATA, ...).
Those constraints also meant native applications instead of Web 2.0 NodeJS and what not. Java specifically had a bad name, and back then not even a coherent UI toolkit.
As always: two steps forward, one step back.
The single biggest problem it has is slow forking. I learned to write my scripts in pure bash as much as possible, or as a composition of streaming executables, and avoid executing an executable per line of input or similar.
As a dependency of a shipping Windows application that needs to cleanly coexist side-by-side with existing Cygwin installations and optionally support silent install/upgrade/uninstall through mechanisms like SCCM, Intune, and Group Policy?
Not so much.
I do use the setup program to build the self-contained Cygwin root that's ultimately bundled into my program's MSI package and installed as a subdirectory of its Program Files directory, however.
If you're curious, I believe the issue was discussed at length in the Go GitHub issues years ago. Also on the mailing lists of many other languages.
> slow forking
There isn't much that can be done about that: starting up and tearing down a process on Windows is much more resource intensive operation than most other OSs because there is a lot going on by default that on other OSs a process ops into, only if it needs to, by interacting with GUI libraries and such. This is why threads were much more popular on Windows: while they are faster than forking on other OSs too, especially of course if data needs to be shared between the tasks because IPC is a lot more expensive than just sharing in-process memory, the difference is not as stark as seen under Windows so the potential difficulties of threaded development wasn't always worth the effort.
Cygwin can't do anything about the cost of forking processes, unfortunately.
https://frippery.org/busybox/index.html
It has a subset of bash implemented on Ash/Dash. Arrays are not supported, but it is quite fast.
The forking problem is still present, though.
For example, someone might do something like this (completely ignoring the need to quote in the interests of illustrating the actual issue, forking):
Instead of something like this: This avoids a sed invocation per loop and eliminates self-renames, but it's harder to work with.Of course the code as written is completely unusuable in the presence of spaces or other weird characters in filenames, do not use this.
Dash has been benchmarked as 4x faster than bash. The bash manpage ends by stating that "bash is too big, and too slow."
To solve the problem or because you saw "slow" and "bash" and wanted to bring up something cool but unrelated?
If I go from 10 seconds of forking and .04 seconds of shell to 10 seconds of forking and .01 seconds of shell, I don't actually care about how cool and fast the shell is. And I've never had the speed of bash itself be a problem.
Bash is prominent in announcing that it is "too big and too slow." It has said this for years. Why are its supporters so firmly in denial?
Optimizations are cool and all, but being 4x faster at something that's already taking negligible time is not something that makes a big difference.
How long ago was that line written anyway? It feels like complaining about emacs using its "8 megabytes" of memory in the modern day.
The solution to this problem is to reduce the number of forks. Faster code on either side is aesthetically nice but unhelpful.
It's not "denial" to acknowledge that an already-fast program has a faster replacement and then get back to working on the bottlenecks.
So did Debian and Ubuntu, so they demoted bash.
Whatever smugness you are interpreting here is diametrically opposed to the facts of this situation, and repeat after me:
It’s too big and too slow.
What do you think 'this situation' is?
Repeat after me: The situation is that fork is using almost all the runtime.
The only solutions are fixing fork or forking less. Your suggestions are not related to the actual problem the user is having.
Java was ahead of its time, now nothing has a coherent UI toolkit.
You don’t get an app that looks the same across platforms. You do get apps that look like they belong on your platform, even though the code is cross-platform. It uses the native toolkit no matter where you run it across Windows, GTK, Qt, Motif, macOS/Carbon, macOS/Cocoa, and X11 with generic widgets.
Older platforms are also supported, like OS/2, Irix, and OSF/1.
https://wiki.wxwidgets.org/Supported_Platforms
It’s a C++ project, but it has bindings for most of the languages you’d use to build an application. Ada? Go? Delphi? Ruby? Python? Rust? Yes, and more. https://wiki.wxwidgets.org/Bindings
So, if you use wxWidgets, you probably have to use either C++ or Python version, others are unlikely to be supported.
Among actively developed bindings, there is also wxRust at https://crates.io/crates/wxdragon
This is false. https://news.ycombinator.com/item?id=24250968 https://news.ycombinator.com/item?id=24259040 It was false in 2020 and it is still false today (I just checked).
I wish the Wx proponents would stop saying these things. Who exactly are you trying to fool? Do you have no concept of reputational damage? What good comes from a claim that is so easily disproven by just installing a Wx application and looking?
I’m not trying to fool anyone. I'm not affiliated with the project. I’m just aware of it and have used it a few times. You, on the other hand, have called me a liar and a fraud because I repeated exactly what the project docs state and which your two links do nothing to contradict. In fact, you linked to yourself being corrected by the actual maintainer of the project. Did you read anything he wrote?
I think I do. I have taken a few minutes on the Web to compare that what I had in mind is correct. What was the point of asking this question? Was it to trap me in a gotcha, or paint me as clueless, or what?
> have called me a liar and a fraud because I repeated exactly what the project docs state
Good, you realise you are taking on the claims made by Wx on paper. However, there's more to the world. To get the full picture, you have to also engage with what I have listed. The docs say one thing, the reality shown in the screenshots say another. There is a contradiction. It remains unresolved, not for lack of trying on my part.
> your two links do nothing to contradict
You are not further allowed by me to invalidate what I was writing about by simply disregarding the evidence. Engage with the points I was making. The differences in look and feel between Wx and native are plain for everyone to see and verify. So, what now? Who is right?
> Did you read anything he wrote?
Yes. Examine this:
his claim> OTOH all the standard UI elements (buttons, checkboxes, text controls, date pickers, ...) are native
my counter-evidence> Well, let's verify that… https://i.imgur.com/uHfjoUs.png No, they're not.
his deflection> Sorry, I don't know what is this supposed to prove
So instead of admitting that there is a contradiction, he just pretends to not understand it.
Also examine this:
> look good
> look good
> looks fine
> look good
I never mentioned anything about looking good, this is a distraction designed to deflect from the central point I was making. As I wrote before, the central point made by me remains completely unaddressed.
Alas, I cannot deal with those crazy-making techniques, his behaviour measured by outcome is indistinguishable from the mentally ill. With the help and advice from a friend, I came to the conclusion that it was not safe for me to respond, so I then decided not to.
I accept that you’ve had a bad experience with the project and that you are speaking in good faith about that experience. I do not accept that I’m a liar or a shill because my experience does not match yours, and you will not convince me to accept that.
Whatever has gone on in other places between you and some third party has nothing to do with me. I didn’t see any credible threat to your safety in the links you shared, but I’m willing to guess that’s not the complete story between the two of you. I really can’t and won’t be dragged into the middle of that, because I don’t know either of you.
I'm sure they did the best they could ... it was just really painful to use.
Maybe so, but my memory of Cygwin was waiting multiple seconds just for the Cygwin CLI prompt to load. It was very slow on my machines.
What I usually do is make sure my code builds with both Cygwin and MingW, and distribute the binaries built with MingW.
Cygwin doesn't work at all in Windows AppContainer package isolation; too many voodoo hacks. MSYS2 uses it to this day, and as a result you can't run any MSYS2 binaries in an AppContainer. Had to take a completely different route for Claude Code sandboxing because of this: Claude Code wants Git for Windows, and Git for Windows distributes MSYS2-built binaries of bash.exe and friends. Truly native Windows builds don't do all the unusual compatibility hacks that cygwin1.dll requires; I found non-MSYS2-built binaries of the same programs all ran fine in AppContainer.
A lot of this is issues Microsoft could fix if they were sufficiently motivated
e.g. Windows lacks a fork() API so cygwin has to emulate it with all these hacks
Well, technically the NT API does have the equivalent of fork, but the Win32 layer (CSRSS.EXE) gets fatally confused by it. Which again is something Microsoft could potentially fix, but I don’t believe it has ever been a priority for them
Similarly, Windows lacks exec(), as in replace the current process with new executable. Windows only supports creating a brand new process, which means a brand new PID. So Cygwin hacks it by keeping its own PID numbers; exec() changes your Windows PID but not your Cygwin PID. Again, something Microsoft arguably could fix if they were motivated
They did fix it, in a sense, with WSL1 picoprocesses. Faster and more compatible than Cygwin. Real fork and exec on the Windows NT kernel. Sadly, WSL2 is even faster and more compatible while being much less interesting. WSL1 was pretty neat, at least, and is still available.
In any event, this diversion doesn't change my analysis of Cygwin. Cygwin still sucks regardless of whose fault it is. I intentionally left this stuff out of my post because I thought it was obvious that Cygwin is working around Windows limitations to hack in POSIX semantics; it's the whole point of the project. None of us can change Windows or Cygwin and they're both ossified from age and lack of attention. We have to live with the options we've actually got.
If you need a Windows build of a Linux tool in 2026 and can't use WSL, try just building it natively (UCRT64, CLANG64, MSVC, your choice) without a compatibility layer. Lots of tools from the Linux ecosystem actually have Windows source compatibility today. Things were different in the 90s when Cygwin was created.
So now even today, compiling any GNU package means probing every last feature from scratch and spitting out obscenely rococo scripts and Makefiles tens of thousands of lines long. We can do better, and have, but damn are there a lot of active codebases out there that still haven't caught up.
It doesn't help that the package simply named "gcc" is for the MSYS2 target.
Instead, you either want UCRT64 or CLANG64, depending on whether you want to build with the GNU or LLVM toolchains, as it uses the newer, fully-supported Universal C Runtime instead.
As for UTF-8 support, it's the manifest file that determines whether Windows sets the ANSI code page to UTF-8. (There's also an undocumented API function that resets the code page for GetACP and the Rtl functions that convert ANSI into Unicode. But this would run after all the other DLLs have finished loading.) Having the code page correct is enough to support Unicode filenames and Unicode text in the GUI.
It just won't provide UTF-8 locale support for the standard C library.
I can certainly relate to this: I'm currently sitting on a request for an enhancement to a product (currently running on a 32-bit Windows 10 VM) with a build system that has never been updated to support any Microsoft platform other than MS-DOS, or toolchain newer than Microsoft C 5.1.
This made me laugh. It reminded me of course work I did in university that was clearly written many years before I took the course as it recommended we manually enable the "new" c99 standard in our compiler, which I guess survived in the documentation up through when I took the course, at which point it was still relevant since GCC was otherwise defaulting to C11 by the time I was using it.
https://github.com/skeeto/w64devkit
I was a Cygwin user from about 1999 to 2022 or so, spent a little time on wsl2 (and it's what I still use on my laptop) but I'm fully Linux on the desktop since last year.
https://en.wikipedia.org/wiki/Hyper-V#Architecture
Requiring every single Linux app developer to recompile their app using Cygwin and account for quirks that it may have is not the correct approach. Having Microsoft handle all of the compatibility concerns scales much better.
Cygwin started in 1995. Microsoft wasn't cooperative with FOSS at all at that point. They were practicing EEE, and eating some expensive Unix/VMS machines with WNT.
Interix was implemented as proper NT kernel "subsystem". It was just another build target for GNU automake, for example.
(Being that Interix was a real kernel subsystem I have this fever dream idea of a text-mode "distribution" of NT running w/o any Win32 subsystem.)
[0] https://en.wikipedia.org/wiki/Interix
Cygwin still lacks that to this day, you have to fire up to GUI installer to update packages.
MSYS2 is cygwin with pacman cli.
Back when I was still using windows (probably XP era), I used to run colinux, it was kind of amazing, setting up something like LAMP stack on the linux side was a lot easier and then using windows editors for editing made for quite nice local dev env, I think! Could even try some of the X11 servers on windows and use a linux desktop on top of windows.
When I noticed I kept inching towards more and more unixy enviornment on the windows, I eventually switched to macOS.
Apart from the obvious hack-value, I can't quite imagine even pretend use-case, with some 486 era machine, you would be limited by memory quite quickly!
[0] http://colinux.org/
What I especially like about this Windows 9x Subsystem project, is that it proves that coLinux could have been written way earlier. Now imagine how less dual booting we would have needed in 1996 if that happened, and how it would have affected VMware which only existed since 1998.
To me, this seems an impossible feat.
But I wonder how it seems to people who understand how it works?
I'm reminded of this joke:
Two mathematicians are talking. One says a theorem is trivial. After two hours of explanation, the other agrees that it is indeed trivial.
In a literal sense, it very well may have been trivial, even if neither you nor the professor would have been able to easily show it.
The one I've always flown with is, trivial means (1) a special case of a more general theory (2) which flattens many of the extra frills and considerations of the general theory and (3) is intuitively clear ("easy") to appreciate and compute.
From this perspective, everything is trivial from the relative perspective of a god. I know of no absolute definition of trivial.
As someone who mostly understands what's going on - It does not seem like wizardry to me, but I am very impressed that the author figured out the long list of arcane details needed to make it work.
There's special hardware in a processor, for the operating system to limit each programs access to memory and processing time, which Windows 9x leaves unused. This means that the Windows 9x Subsystem for Linux can say "look at me i'm the operating system now" and take over that hardware to run a modern operating system.
Windows 3.11 was a hypervisor running virtual machines. The 16-bit Windows virtual machine (within which everything was cooperatively multitasking), the 32-bit headless VM that ran 32-bit drivers, and any number of V86 DOS virtual machines.
Win9x was similar in the sense that it had the Windows virtual machine running 32-bit and 16-bit Windows software along with V86 DOS VMs. It did some bananas things by having KERNEL, USER, and GDI "thunk" between the environments to not just let 16-bit programs run but let them continue interacting with 32-bit programs. So no, Win9x was in fact 32-bit protected mode with pre-emptive multitasking.
What Win9x prioritized was compatibility. That meant it supported old 16-bit drivers and DOS TSRs among other things. It also did not have any of the modern notions of security or protection. Any program could read any other program's memory or inject code into it. As you might expect a combination of awful DOS drivers and constant 3rd party code injection was not a recipe for stability even absent bad intentions or incompetence.
Windows 2000/XP went further and degraded the original Windows NT design by pulling stuff into kernel mode for performance. GDI and the Window Manager were all kernel mode - see the many many security vulnerabilities resulting from that.
WSL9x uses the same Win9x memory protection APIs to set up the mappings for Linux processes, and the memory protection in this context is solid. The difference is simply that there is no need to subvert it for compatibility.
What Windows 9x didn't have was security. A program could interfere with these mechanisms, but usually only if it was designed to do that, not as a result of a random bug (if the entire machine crashed, it was usually because of a buggy driver).
Have the model spit out example programs to study the API
What makes you think so?
> Proudly written without AI.
> Proudly written without AI.
It’ll produce something, sure, but it won’t actually work, and making it work takes as much effort as building it from scratch.
In the math space it's not even quite as silly as it sounds. Something can be both "obvious" and "true", but it can take some substantial analysis to make sure the obvious thing is true by hitting it with the corner cases and possibly exceptions. There is a long history of obvious-yet-false statements. It's also completely sensible for something to be trivially true, yet be worth some substantial analysis to be sure that it really is true, because there's also a history of trivial-yet-false statements.
I could analogize it in our space to "code so simple it is obviously bug free" [1]... even code that is so simple that it is obviously bug free could still stand to be analyzed for bugs. If it stands up to that analysis, it is still "so simple it is obviously bug free"... but that doesn't mean you couldn't spend hours carefully verifying that, especially if you were deeply dependent on it for some reason.
Heck I've got a non-trivial number of unit tests that arguably fit that classification, making sure that the code that is so simple it is bug free really is... because it's distressing how many times I've discovered I was wrong about that.
[1]: In reference to Tony Hoare's "There are two ways to write code: write code so simple there are obviously no bugs in it, or write code so complex that there are no obvious bugs in it."
That makes sense, I was just born yesterday.
Unfortunately this is ambiguous, as there's an AI product called Zero AI.
Looks like it's been updated now to be more clear. Amazing though.
note: i'm not saying author did not improve his skills overall, but also last '6 years' perhaps also means - fair amount of digging the web with search engines, which are... like AI 0.1
https://github.com/mastodon/mastodon/issues/23153
https://github.com/mastodon/mastodon/issues/19953
https://github.com/jwilk/zygolophodon
I've been working on a WebExtension that calls out to zygolophodon and returns plain HTML to the browser. In the process of rebasing it over recent changes but here is the working webext-old branch:
https://github.com/jwilk/zygolophodon/compare/master...pabs3...
Yes, I have weird problems. I get to look after some very weird shit.
Still got those in this part of the world sharing space with state of the art autonomous 100+ tonne robo trucks.
I actually built a win9x compatibility mode into BrowserBox specifically for this kind of weirdness. You run the server on a modern system and launch it with bbx win9x-run, and it proxies the modern web to legacy clients. It works surprisingly well with IE5, IE6, and old Netscape on Windows 95/98/NT. Might be a fun addition to your retro utility belt!
Or better, ditch the web completely and head to Gopher/Gemini.
I get the degradation to Gopher as a way to solve many issues, but many things just don't work there. And it may have its own vulnerabilities.
Run it in Windows XP, in a VM.
Now here's the clever bit - qemu will allow you to expose the keyboard, mouse, and framebuffer as a VNC server. So you set up Apache Guacamole to point a VNC client at the VM, and then "normal people" can log in, operate the transmitter, and log out again.
You can do a lot of sneaky things with that, including setting up headless X, running VNC on it pointed at your qemu VM, and then streaming the headless X servers's framebuffer out with ffmpeg.
Yes sometimes work can be a bit boring with not much to do, why do you ask?
BrowserBox is basically the same pattern as this setup (streaming graphics from some browsing substrate somewhere to web clients) except architecture is different: a modern box on the same private network runs the BrowserBox server, and the win box (QEMUd or otherwise) connects to its http endpoint, using whatever browser it has (tested back to IE5 even, tho that's a way more buggy browser than IE6. IE8 should be golden). That way you get the full modern web, no compromises. But crucially the web is not actually accessing your legacy box. So, no comrp0mises, ie., no easy vulns. Especially a concern for older browsers. Plus, we've got policy controls to lock down capabilities (copy, paste, URL lists, internal IP access controls, etc).
In your case it sounds like you are running the webby servers on the XP box, too, so BrowserBox would link back into those over the same private network, render them on the modern box, send it back to the XP box, then clients can connect over the QEMU VNC bridge you already have.
Alternately you could just do away with the Win XP browsing, and have BrowserBox connect to your webby endpoints for transmitters wherever you run them, and then expose that browsing graphics stream to clients over whatever endpoint you want. Many options!
I like your ffmpeg out setup. How did that go? Share more about that? Pretty interesting, I love this old architectures, and legacy systems compatibility quests.
> In your case it sounds like you are running the webby servers on the XP box, too,
Yeah - it's a web front end to some specialised software written on I guess Microsoft C++ (if I had time, enthusiasm, and a copy of it lying around I suppose I'd wave Ghidra at it and see what happens).
I'll look into BrowserBox, that sounds handy.
> I like your ffmpeg out setup. How did that go? Share more about that? Pretty interesting, I love this old architectures, and legacy systems compatibility quests.
Surprisingly well. I think you could probably stream it straight to Twitch or something if you wanted. Yeah, this sounds like a blog post.
There is a section of the Forties Pipeline where they have a huge amount of gas handling plant in central Scotland. Last time I was on site (admittedly 15 years ago but I don't see this changing soon) the SCADA outstations were run by absolutely minty box fresh VAXStation 3100s. Plastic not even peeled off the front panel badges fresh.
Just few months ago seen windows 95 error message on HSBC ATM.
I recently saw that running on special 16 channel DAT recorders used by the 999 service, recently as in "within the past five years". I believe they've been retired but kept around in case they need to recover tapes off them.
I kept my mouth ABSOLUTELY FUCKING SHUT about knowing my way round OS/2 Warp 4.
Except it would be nice if computers were as snappy as back then... so many layers of cruft that nothing really pops. I remember I used to just disable all the NT/2K animations and it felt so nice. I built a career on web apps, but most are just really poorly written it pisses me off.
You should try Haiku: https://www.haiku-os.org/
It's actually pretty usable these days, at least on Thinkpads and other "generic Intel chipset" machines.
I wonder how similar this project is to "BSD on Windows": https://archive.org/details/bsd-on-windows
Also, I know about https://en.wikipedia.org/wiki/Architecture_of_Windows_9x, but it's not really meaty enough for my taste. :)
It's got lots of very thorough documentation and sample code to dig through
Other well known anecdotes and pieces:
https://jacobfilipp.com/msj-index/
https://web.archive.org/web/20240318233231/https://bytepoint...
(Edited: mixed it up on the last sentence.)
Which is exactly what the post says this is. It's running Windows 9x on Linux kernel. It's strangely worded, but from the follow up comment, and the readme in the repo it seems clear that it's running on the Linux kernel.
Pretty sure it's the other way around. But I haven't had my coffee yet ;)
Microsoft's naming scheme confuses me every single time though: "Windows Subsystem for Linux" actually runs Linux on Windows...
I missed the part that it runs on the bloody Windows 9x kernel, I was to busy thinking about modern Windows.
Per wongarsu's post, something like the OS/2 Subsystem is an OS/2 system with Windows beneath it, but the OS/2 Subsystem is much smaller and less consequential, thus subsidiary (in the auxiliary sense) to Windows as a whole.
Isn't marketing fun?
This is how we end up with hundreds of products that provide "solutions" to your business problems and "retain customers" and upwards of a dozen other similar phrases they all slather on their frontpages, even though one is a distributed database, one is a metrics analysis system, one handles usage-based billing, one is a consulting service, one is a hosted provider for authentication... so frustrating trying to figure out just what a product is sometimes with naming conventions that make "Windows Subsystem for Linux" look like a paragon of clarity. At least "Linux" was directly referenced and it wasn't Windows Subsystem for Alternate Binary Formats or something.
WSL9x takes quite a different approach. Windows boots first, but once Linux starts both kernels are running side-by-side in ring 0 with full privileges. They are supposed to cooperate, but if either crashes then both go down.
These days especially, it's better to just use 86Box for those operating systems.
But CoLinux - IIRC - required the NT branch of Windows. I can only imagine the level of hackery it takes to make this happen on Windows 9x.
Part of me wants to weep at the sheer perversity, part of me wants to burst into manic laughter. It is indeed a world of endless wonders.
I have not tested this yet, so I have no idea - but if he managed to pull this off then this may be one of the greatest achievements this year. Or perhaps there are some restrictions to it? Does compiling stuff work in it? So many questions ... who has the answers?
Maybe there's some detail I don't quite follow, like is has support for 486, but only those with a built in FPU?
> Because we cannot name something leading with a trademark owned by someone else.
https://xcancel.com/richturn_ms/status/1245481405947076610?s...
And this WSL project is going to run into the same problem.
The "for Linux" is added because it's a subsystem for Linux applications (originally not leveraging a VM).
Microsoft also had the "Microsoft POSIX subsystem" (1993) and "Windows Services for UNIX" (1999) which were built on the "Subsystem for Unix-based Applications" (rather than "Unix-based Application Subsystem"). That chain of subsystems died at the end of Windows 8, though.
There are many reasons not to put "Linux" in front, but the naming is consistent with Microsoft's naming inconsistencies. It's not the first time they used "subsystem for" and it's not the first time they used "Windows x for y" either.
The naming is ambiguous, you could interpret the Windows subsystem for Linux as a subsystem of Linux (if it had such a thing) that runs Windows, or as a Windows subsystem for use with Linux. Swapping the order doesn't change that.
In other languages, the difference would be clearer.
I do agree it's an issue of English being an imprecise language.
And this is a poor example, because Microsoft wants to be Microsoft.
https://en.wikipedia.org/wiki/Windows_NT_3.1#Architecture
It can work either way though.
The name we shipped was even worse than Windows Subsystem for Linux, honestly. At least Microsoft spent some time on it.
OpenOffice XML [1] -> Office Open XML [2]
[1] https://www.openoffice.org/xml/general.html
[2] https://en.wikipedia.org/wiki/Office_Open_XML
Do any screen editors work in the command prompt windows? Try with "export TERM=ansi".
Windows 9x, by contrast, was DOS-derived. Running Linux inside it would require fundamentally different (and messier) hacks - which is probably why nobody did it at the time. The very fact that this works at all is a testament to how ahead-of-its-time NT's architecture was.
For a practical answer: you'd need something like this for legacy locked-in situations. Old medical or industrial software that only runs on Windows 98, or specialized hardware without modern drivers. That said, if you have a 486 handy in 2026, running Linux natively is almost certainly more useful than running it inside a 30-year-old DOS derivative.
Some interesting reading:
1. What was the role of MS-DOS in Windows 95? (https://devblogs.microsoft.com/oldnewthing/20071224-00/?p=24...)
2. Why doesn’t Windows 95 format floppy disks smoothly? (https://devblogs.microsoft.com/oldnewthing/20090102-00/?p=19...)
3. Running MS-DOS programs in a window alongside other Windows programs (https://devblogs.microsoft.com/oldnewthing/20231127-00/?p=10...)
If you want to run your windows software in Linux, you could try Wine. Wine seems to have support for WNASPI so it's possible your software would just work. (You might have to run Wine as root I guess, to get access to the SCSI devices.)
If Wine doesn't work, Windows in QEMU with PCI passthrough to the SCSI controller might have better chances to work.
Wines WNASPI32.dll is really just a facade - it doesn't provide actual SCSI services, its just there for SCSI-using apps to think they have ASPI onboard - so for my case I would need to write a shim to pass through SCSI IO requests to a Linux service - or loopback file? - to actually process the requests. I've been meaning to do this for a long time, but if there is some way I can set up a loopback file under Linux to 'pretend' to be a SCSI block device for a Windows app, I'd sure like to know if its possible ..
(To clarify: while the kernel is actually running at ring 0, to act as a driver it seems to use the usermode profile.)
The Wikipedia page is not verify informative and presents it as a regular VM (possibly mixing up 9x and later versions that run the NT line of kernels). The manual is a bit more informative about the tech:
ilab.usc.edu/packages/special/Win4Lin-3.0.1-manual.pdf
I’m a bit surprised it hasn’t been mentioned a lot in the comments. Maybe it’s a bit too old for most people here (Linux in the late 90ies/early 00s was a much smaller community)?
Does it live there irrespective of this project? Or is that part of the patching?
> looks inside
> virtual 8086 mode
This trickery is called binfmt_misc , which is a linux kernel system to associate random binary files with custom userspace 'interpreters'
I have had it working in the past. And while it is kinda neat I prefer manually running 'wine program.exe' to have a bit more control.
I have seen reports that a binfmt_misc setup + wine is good enough to get infected by certain windows viruses ;-P
To get around that, I recently added a legacy compatibility mode to BrowserBox (bbx win9x-run). It basically lets you run the server on your modern daily driver, and access it via IE 5, IE 6, or Netscape on the Pentium box. It strips away the modern TLS/JS rendering issues and lets you actually browse the modern web from Windows 9x. Highly recommend giving it a spin if you get that machine built!
Edit: to the people who downvoted, I corrected my spelling mistake. I will perform due penance.
I built a Win9x compatibility mode for BrowserBox that does exactly this (https://github.com/BrowserBox/BrowserBox/blob/main/readme-fi...). Ur modern server does all the rendering, and it outputs a client link specifically designed for legacy browsers like IE5, IE6, and Netscape running on Windows 95/98/NT, streaming them the pixels. It's definitely an abomination, but there's something magical and retro that I like about viewing the 2026 internet through an IE6 window ;) ;p xx
- Retrozilla with some about:config flags disabling old SSL cyphers and new keys to enable newer ones
- Iron TCL maybe with KernelEx and BFGXP from https://luxferre.top reading gopher and gemini sites such as gemini://gemi.dev proxying all the web bloat and slimming it down like crazy
- Same Gemini URL, but thru http://portal.mozz.us/gemini . Double proxy in the end, but it will be readable.
I mean it's like trying to balance a cybetruck into 4 skateboards and flunging it over a hill cool