Hacker News Reader

Comment by greenn | Hacker News Reader

I haven't profiled our Windows kernel driver across different kernel versions (maybe I should!) but I'd like to offer this perspective: the kernel is incredibly stable from my driver development point of view. The biggest reason we ship different drivers for Windows 7/8/10 is just that newer WDKs don't support anything older than 10. The kernel has remained remarkably consistent while still offering new features we can take advantage of on non-legacy systems.

View on HN

I don't know why "incredibly stable" is such a remarkable thing, the whole point of an operating system and a kernel is to offer a stable API to write your applications against.

I'm talking about the kernel space itself, not the APIs exposed to userland to interface with the kernel from your application. Internal APIs and behaviors are mostly identical over the past ~20 years, and any changes are usually moved to a new export. I don't think this should be taken for granted.

>I don't know why "incredibly stable" is such a remarkable thing

It's remarkable when you look at the landscape of Linux and Mac device drivers.

Can you run non-kernel drivers for Linux 2.6 on 6.6? Can you install a device driver from 2007 on a modern MacOS? Well, many Windows 7 drivers also work on 11. That's stability.

That's kind of the point I'm trying to make here, an operating system and kernel is supposed to provide you with something stable to build on. It's its raison d'être. Should the stability of a bridge be considered remarkable? No, it's an entire engineering discipline to making stable bridges, because an unstable bridge is useless. The same goes for an operating system and kernel.

> Should the stability of a bridge be considered remarkable?

From a purely engineering perspective, most certainly.

If it isn't so remarkable why doesn't every OS do it?

I can use Windows drivers from Windows XP era. Try that on Linux.

And I can run any win32 binary, regardless how old is it. Try that on Linux.

I've often read that, but in my experience it's not true. Drivers before 7 don't work on 7+. Exes from win XP era often fail to run, even with the compatibility modes offered by the OS. Heck, the only times I used compatibility modes where as workarounds for binaries that didn't exist when XP support had already ended...

I thought that Windows 10+ after a certain build made signed drivers mandatory without safe mode or other workarounds? I know I’ve tried Windows 7 drivers for some things and ran into that issue.

A hello world driver maybe? My XP printer driver didn't work on 7, had to buy a new one.

> I can use Windows drivers from Windows XP era. Try that on Linux.

Linux has a model where all drivers should live in-tree; if we account for that, then yes, most devices that worked on Linux in 2001 will work on Linux today.

> And I can run any win32 binary, regardless how old is it. Try that on Linux.

Yes, Linux also has excellent compatibility with old win32 binaries. This is partially a joke and partially not.

I think the reason you're finding it to be so stable is because nothing in that kernel has changed since Windows XP.

A fair amount has been changed or added, I'd say. I recommend the Windows Internals book if you're interested in learning about the Windows kernel. The 7th edition just came out.

That's fair. My comment is not well-informed with regard to whatever has happened in the last 10 years.

The impression I formed back in the late-2000s and early 2010s when I was doing this type of development was that there was very little change in the low-level NT kernel. I actually don't think this is a bad thing; going from XP to Vista was a breeze because nothing about the driver APIs changed in the slightest, and Microsoft even provided a bunch of new filesystem examples with the Vista DDK.

Oh you're totally right about the driver APIs. A fair amount has been added, but that stuff gets a new export to avoid breaking anything. For example, there's a dozen different `ExAllocate`s now. I wouldn't call myself a driver-writing expert but I'm sure that same driver for Vista could be made to target Windows 10 with little effort.

That's just plain wrong.

Reply on news.ycombinator.com