Hacker News Reader

Comment by BuildTheRobots | Hacker News Reader

> Hackers reverse engineer Microsoft SQL server for Linux compatibility DLLs and adds better Wine support. Microsoft will eventually start selling a commercial Windows 10 compatibility layer for Linux

I like the optimism but I honestly believe it's going to go the other way round.

Based on Microsoft s EEE strategy from the past, combined with the way they've pushed for SecureBoot on x86 and Arm and the (re-)rise of the WSL, I honestly expect it to get even more difficult in the future to boot a linux kernel on your own hardware.

Once all the usual suspects work properly under WSL I think more hardware lockdowns are going to come into place combined with phrases like "what do you need a kernel for? just boot windows kernel and use your linux userland and stop complaining!". The rise of GNU/Windows might well be coming...

View on HN

I think Microsoft was right to push for Secure Boot / bootloader signing. The way they pushed for it was/is anti-competitive.

I don't think UEFI was the right horse to put money behind. It's a complicated hot mess and the most visible innovation has been to use UEFI identifiers for DRM within web browsers. Yay.

With intel ME next to it, it seems like asking for a metal door, with special key shapes the competition has a hard time to produce, but on a wall you know you can break with a hammer.

The current state of computing is less and less about being pro people, and more and more about pro big players.

>I think Microsoft was right to push for Secure Boot / bootloader signing

Signing by whom? The OEM of course because giving the key to the user would be unacceptably user-friendly for microsoft.

>I don't think UEFI was the right horse to put money behind. It's a complicated hot mess and the most visible innovation has been to use UEFI identifiers for DRM within web browsers. Yay.

UEFI was done that way for a reason. Don't underestimate Microsoft's influence on other companies. Same thing with big web companies pushing "open" standards that only they have the manpower to comply with.

Or systemd/Windows, depending on how long it takes.

The only thing missing from SystemD to be a worse Windows is a binary configuration file.

That would be wonderful. If you're comparing kernel architectures, Linux kernel vs. NT is not much of a comparison. I will take NT kernel & gnu user space any day!

It's always nice to run a kernel designed by people who actually know what they're doing, and aren't doing it for the first time. Look into nt history, or how late Linux kernel started taking things like power management seriously. Hell, Linux kernel still insists on entirely living in nonpageable memory, most io is still synchronous, and there still is no stable driver API.

The linux kernel ain't perfect, but to say the NT kernel is superior is laughable imho. Take it with a grain of salt though as I am a greybeard sysadmin type who got so fed up with windows and MS that I moved completely to gnu/linux and haven't looked back.

Back on topic though, Skyrim normal edition installed and played just fine with proton for me on manjaro. This is nothing but good for us linux gamers.

Windows of today is nothing like what the original NT architecture was designed like.

NT sends CPU chasin' pointers between layers and layers of (filter) drivers in IRP madness.

But yeah, it's so great on NT I can set that 16 kB of IRQL passive level driver code pageable. Massive savings, and I'll only need to make sure I never ever run any of that code on IRQL dispatch or higher. If I mess up, the end user has a pretty blue screen to stare at.

I guess the impact of a pageable kernel was more important 10-20 years ago. Nowadays I rather allocate nonpageable memory just to avoid the headaches of not being able to actually use that memory at typical IRQLs a driver requires. For example DPCs run at dispatch IRQL, no access to pageable memory.

Reply on news.ycombinator.com