Microsoft contributes a lot of changes to Linux kernel 3.0 | Hacker News Reader

Microsoft contributes a lot of changes to Linux kernel 3.0

h-online.com

111 pointsyarapavan33 comments

While I would agree (and some would make the point) that Microsoft does this to benefit mostly Microsoft, it's still GPL'ed code that is now part of the greater Linux kernel codebase, so good on them and everyone else.

Open source is most effective when multiple parties find their self-interest overlaps.

Embrace and extend, the contribution is basically VM support. Run Linux in a VM, just like XP-mode in win7. The SMB team or ODF people would never go cheering about MS's self-interest...

> Microsoft does this to benefit mostly Microsoft

Well, of course. I would wager all the corporate contributions, and a good chunk of the independent contributions are contributing to add features they need.

I agree, though, good on them for contributing, and even better for making a substantial contribution, instead of a token gesture.

The irony is, there is a benefit to Microsoft for working with open source: if dosbox and wine become good enough to support their legacy software, they can end-of-life a large portion of their compatability codebase that they drag along with each version, and keep the compatability effort to a stable, free platform written for that purpose.

Which product are you talking about?

They don't need dosbox/wine -- "XP mode" in windows 7 is XP running in a virtualized environment.

But their marketing had the brilliant idea of making XP mode only work on some Windows 7 licenses.

Ah, but how does that replace DOSBox? iirc, Windows XP does not support real mode DOS software.

It does. http://en.wikipedia.org/wiki/Virtual_DOS_machine

NTVDM is unavailable in 64-bit Windows, but XP Mode uses 32-bit XP, so it can be used to run DOS software. (Though there will be trouble if you try to use it to run DOS games.)

NTVDM is useful, but only for the most basic software. For example, many programs do not work on NTVDM because of the limited audio and video support. You mentioned games, and I believe some of them suffer from timing issues (though this may be the games' fault for relying on a specific clock frequency for timing).

I'd say it's just good enough for the most useful 16-bit Microsoft utilities.

I've seen the entire suite of DOS Borland C++ developer tools run under NTVDM, and other non-trivial, useful stuff, like Lotus Magellan.

I've also seen some quite trivial programs fail completely in DOSBox due to imperfect compatibility.

Huh? Microsoft would never do that. Tell their customers to use Free/OSS system? An MS controlled VM system is exactly what they want, and exactly what they've done.

Microsoft have so far not advocated competing systems.

The initial release of the iPod only interfaced to a Apple computer.

Oracle makes a lot of money.

Microsoft likes money.

At some point being more open may be a Microsoft strategy.

Certainly I think they would be more successful if they were more open.

Well, IBM's contributions to Linux are to benefit IBM - they make their money on hardware and services, Linux is what we call a "loss-leader" for them. Red Hat's contributions to Linux are to benefit Red Hat - they make their money on selling support contracts, they've got to have something to support! Intel make their money on CPUs, of course they want to tweak the OS...

Basically I don't think that enlightened self-interest is a bad thing.

Not really, MS still claims a patent against Linux used to bash Android OEMs ..accepting this means that we have still not forced MS into a patent lawsuit allowing us to invalidate the patents..

In my experience, people will always find a way to pooh-pooh the open source contributions of organizations they don't like, e.g. Apple and WebKit/Darwin.

Looking at the stats, it's interesting the Google are so low (1.4% of changesets) when they do so much work on both Android and Chrome OS. Does their work not go upstream?

It's also interesting that Nokia are at 1.8% of changesets when Elop siad they weren't producing any more MeeGo phones.

Is most work that goes into Linux now just adding drivers for new hardware? Or are there ineresting new kernel features that people are working on?

From what I recall something like 50% of the code is for device drivers, though there is interesting stuff going elsewhere. LWN covers it pretty well (in fact nowhere else really does at all) http://lwn.net/Kernel/

Also, those stats don't take into how used something is. You can dump 200k lines for a device only used by 0.1% of users, yet if you wrote something of 2k lines, used by every user (scheduling/filesystem etc.) you wouldn't make the table, even though your impact is greater.

48.8% by size in the drivers directory:

linux-3.0-rc7$ du -sc * |sort -nr |head -2 505508 total 246524 drivers

> Does [Google's] work not go upstream?

A lot of it doesn't, actually. Many of the Android patches are pretty intrusive, and wouldn't necessarily be suitable for integration into the mainline kernel -- for instance, the Android permissions model depends on a bunch of these kernel patches.

I'm pretty sure that they'd ultimately accept kernel patches for the actual end-to-end functionality that Android needs, but there would be a lot of debate and probably considerable engineering changes to kernel interfaces needed first.

In practice, Android development seems to be more focused on shipping rather than on code quality, accruing technical debt in the process. To be fair to them, it's a project that is really an entire Linux distribution, requires lots of components written from scratch and not used anywhere else, and they're under massive pressure to keep up with the competition.

It seems a shame than Android is so different from standard Linux distributions and doesn't really use any of the normal Linux stack. It would be really cool if Android was more "Linux" than just the kernel (like Web OS & MeeGo are).

If it was, they wouldn't have the tight integration that makes Android vastly more viable than MeeGo.

What does that mean, exactly?

This is interesting in the context of Microsoft patent-trolling companies which deploy Linux (TomTom, Android OEMs).

Is it possible these patches introduce more patent mines into the Linux source?

Or is there a legal hook to prevent Microsoft from trolling after they (even when only represented by a single employee) participated in the development? (I wish there was, but I fear the answer will be "no")

Original LWN article: http://lwn.net/SubscriberLink/451243/060f24c5b1202513/

They only thing they are contributing is their hv driver to make Linux run on Mircosoft HyperV. HyperV has to support Linux. And the major hv developer is a former Novell employee.

While this is true (at least the part I can verify, no idea about personal things here), I'm genuinely curious why you want to point this out?

The article says the same thing (contributing to the HyperV driver). Why is the job history of this guy of interest?

Because it's not Microsoft who is contributing. They just paying Linux Kernel developers to build a driver for their HyperV. So, Microsoft does not contribute to Linux. They make it just run on HyperV because some customers need this "feature".

Isn't this essentially the same way every company is contributing to Linux?

> Because it's not Microsoft who is contributing. They just paying Linux Kernel developers

So it's Microsoft employees as directed by their managers ... but not Microsoft? What is the distinction.

The distinction is "He doesn't like Microsoft, so he must attack and minimize their contributions to feel better."

two reasons. first, because of the historical animosity between ms and linux. second, because the contributions have appeared high in some site's "ranking" of contributions to linux 3.0. part of the article explains (poorly) that this is because the particular ranking was by number of commits, rather than significance (or number of lines changed). this isn't a criticism of the commits (linux prefers many small changes to one large one), but rather the ranking method used, which misleadingly implied that ms was making "major" contributions to linux.

so the whole conversation is addressing a discussion elsewhere that is based on those two points: ms and linux are traditionally "enemies"; "ms" was misleadingly represented as making major contributions to linux 3.

Reply on news.ycombinator.com