Hacker News Reader

Comment by orf | Hacker News Reader

Suggesting a rewrite of something as important as the NT kernel is ignorant at best.

I think Bash on Windows is a great compromise.

Why is the NT kernel so important? It hasn't touched my life much in about eight years, private or professional. Nokia thought Symbian was very important too.

Why switch from a tested kernel that is technically very nice? Nothing about the kernel prevents proper support of POSIX and it has a lot more drivers and hardware support than any existing kernel.

Your exposure to the kernel in the last 8 years is not representative of 90+% of computer users.

Unless you're OK with breaking compatibility with all existing Windows apps - most of which you have no control over - rewriting the kernel as a POSIX thing is going to look a lot like 'bash on windows' in reverse. You'd have to have some kind of shim layer to translate all the old system calls, so most everyone's existing software is going to be slower and there will be bugs, etc. Including your own, since Windows is a kernel, a ton of drivers, a userland (including basic runtimes), a windowing system, a basic set of productivity tools and games, etc.

I think this would be a total disaster that killed Windows, personally.

> 'bash on windows' in reverse

That would be "wine", which has worked surprisingly well several years.

Wine is fine for getting around an expected limitation, but would you run your workloads on wine in a production datacenter? It would be a disaster.

I wonder what companies really have designed new backends for windows for years anymore? Technically you could run many generic things there (databases or JVM), but what would you actually gain? Likely if you have to use windows, you're left in some legacy niche, looking backwards and not into the future.

I'd run my entire development stack on Bash on Windows though. That's what it's there for. It's glorious!

Similarly, my laptop probably wouldn't be able to handle a production workload for any length of time...but it doesn't have to either.

The kernel isn't that import but all the software that depends on it and the win32 libs is. There is lots of software that is w32 only, more then enough to keep windows around for years.

There are thousands, probably millions of small business critical pieces of code running on systems all over the world running more than just .NET code. You'll have to emulate or recreate nearly the entire operating system, at which point... what did you accomplish? Businesses are going to compare the cost/benefits of upgrading their OS vs upgrading some old code, and decide it's not worth it.

Perhaps you've been lucky enough to move to more modern systems, but you're in the tech circle. For everyone else, it's not an easy option. Have you ever worked for a fortune 500 company? There are systems that are nearly 20 years old with billions of dollars running through them. Moving that is not a light decision.

That's the thing. There are many things that are really important that don't actually touch every single person on the planet.

One of the most widely used OS's ever is called Tron. Does that effect me? Not really. Is it important? I think so. I thought so at least enough to do some reading on the subject. Was it important from a CoSci standpoint? Maybe. NT is important from multiple angles, as well. Also, it's importance doesn't take away importance from the things that may actually touch your life. Such is the nature of things.

There's nothing about the NT kernel that needs to be rewritten to implement a POSIX or SUS compliant user-mode subsystem. They once had a Windows Services for UNIX (SFU) subsystem that was POSIX compliant. The new "Bash on Windows", officially called "Windows Subsystem for Linux", appears to be targeting the Linux Standard Base (LSB) and not striving for POSIX or SUS compliance.

LSB is mostly a superset of POSIX. POSIX is a subset of SUS. OSX is SUS compliant, though not POSIX certified. Windows is POSIX certified though not SUS compliant. And no one is LSB certified except a few Linux Distros.

This leaves me wondering exactly what target Microsoft should be aiming for since everyone seems to be playing by their own rules. It's like playing Monopoly with your friend's family who has their own house rules that aren't really like your family's house rules both of which are incompatible with the official rules.

You mention a target as compliance or certification, but I don't think this is a direct concern for the WSL. They want to support specific use-cases, at least at first.

---

"Moving forward we will be investigating new areas of interest while continuing to support the following scenarios:

+ GNU command line tools such as grep, sed, and awk

+ File system and symlink support within the WSL environment

+ Ability to run apt / apt-get for installs, updates, and package testing

+ Basic functionality for languages such as NodeJS/npm, Python, Perl

+ Command line tools such as Vim, Emacs, Git, and SSH"

https://blogs.msdn.microsoft.com/wsl/2016/07/08/bash-on-ubun...

I don't think it's a direct concern either. Though if I had to guess they'd target LSB over any of the others.

I think it's telling that they called it the "Windows Subsystem for Linux" where are their prior Subsystems (SUA/SFU) were for "Unix".

I think software is going to the direction of emulation anyway.

At this point you have platforms like Electron, which run in a browser environment. Java software which runs inside JVM. Now there's even linux sub-system inside windows, etc.

I think there is a way to do it the other way around. MS put effort in creating something similar to WINE and make a special "Windows POSIX", which let's say for the next decade is released along with the classic Windows ( Same as NT before ). Where they try to emulate / rewrite the software piece by piece until the stability is reached to switch.

I don't see this as an impossible idea to me. Actually if they did something like this I will be so happy to switch back.

I don't think your kernel of choice is of direct interest here; I'm fairly certain what you're interested in is access to your familiar tools, which the "Bash on Ubuntu on Windows" system would potentially provide.

As for a POSIX sub-system on Windows, well there once existed such a system that was near-compliant, but MS stopped supporting it a few years back. As for Wine, I think cygwin fulfills a similar role.

Anyway, you want your tools. You might even want your familiar filesystem and its hierarchy. I think the primary reason Microsoft deemed it worth it to implement Bash on Ubuntu on Windows was to provide webdevs access to the same tools they had on osx/linux.

All in all, I think a bit of reading on these subjects would help, because I think you're confusing the kernel with the userland.

Note: I'm faily young and inexperienced, and when I heard the "Bash on Ubuntu on Windows" announcement, I didn't really know what the meant. But learning about these bits on wikipedia/various blogs was certainly interesting.

More significantly, the underlying technology beneath Linux on Windows, especially if it can be made a generally available, documented target for development, is a great thing. Not even a compromise, I think it's potentially better than making Windows into yet-another-POSIX kernel.

Reply on news.ycombinator.com