Microsoft's New Core OS Team Learning from Linux

sokk writes "Seems like Microsoft is paying attention to the Linux way of doing things. According to itworld.com, a new central engineering division will work on the core of Windows: "The Windows Core Operating System Division (COSD), within the company's Platforms Group, will be responsible for the core OS platform, including development, program management and testing, Microsoft said in a statement sent via e-mail.". A little further down the page analyst Rob Enderle: "They have been studying Linux extensively. Part of their study has been on how Linux has been able to maintain a high level of consistency in the kernel while groups around it maintain maximum flexibility,". "

Comments

Ms

[the_real_rs]

Maybe someday we can all work together and learn from each other. Linux got a few ideas from windows. and windows from linux. Hey if windows can be more stable and work more for the user, more power to Microsoft.

Interesting

[fw3]

"creating a new central engineering division"

Microsoft is going to become more centralized to better compete with a competitor based highly distributed, decentralized development.

I'm amused, of course the proof will be in the bits.

but but ...

[Anonymous Coward]

According to the Microsoft apologists and mr Gates himself there is no innovation in Linux...

Soon to come: a new development process invented by and patented by Microsoft.

Funny thing is ...

[WindBourne]

This is the way that NT started. Dave had total control over the core (and the people) while Bill had control over the API on up. Then Bill took back control of all of it. Bill should have left Cutler in control of the core.

take this a little bit further

[BigGerman]

So you got a bunch of MS core people who are _encouraged_ to look at the Linux kernel.
They got some really bright people there you know.
And what that bunch of bright hackers would want to do pretty soon? Contribute. Just like that Microsoft fellow from the original Haloween documents described - he had the urge to make changes, to improve the open-source code he looked at.
Even it they are prohibited from doing so (which I doubt - engineers and lawyers don't mix), the certain cross-contamination of the ideas is bound to happen.
This will lead to the whole bunch of interesting things - from super-SCO-sized legal battle to Longhorn Linux.

"They have been studying Linux extensively..."

[Kevin Burtch]

"Part of their study has been on how Linux has been able to maintain a high level of consistency in the kernel while groups around it maintain maximum flexibility,"

MS, here's a clue: Stop using undocumented/proprietary hooks into your OS from your apps.

Linux is the way it is (in reference to the above quote) because people stick to the "API"... partly because there's no other way, but that's another topic/philosophy alltogether.

the assimilation of methods

[TeamLive]

Microsoft is at another point in its company evolution where it sees a good idea and assimilates it into it's business and product strategy. It is unsuprising, given the fact that the company has in the past profitted from the innovations of others by refining and mass marketing them (i.e. Windows came from observing the innovation of Apple and Xerox's PARC, Internet Explorer came from the observing of Netscape, and the continuing development of PocketPC which came from the observation of Palm.) One can now only hope that they actually implement the philosophy of open source development that has made Linux one of the most robust and flexible operating systems, or whether they simply copy features.... time will tell.

Rob Enderle is not a reliable source

[gvc]

Rob Enderle makes more media pronouncements than just about any other analyst. In general he has no expertise and no information on which to base his statements.

I would completely discount any report that uses him as a source.

Don't take my word for it. Use Google and judge his veracity and competence for yourself.

Re:Rob Enderle is not a reliable source

[bartash]

Wow, thanks for the tip, that was really interesting.

I like this quote:

One issue is the Unix roots in Mac OS X, which is based on the BSD operating system. "This Unix component is working against them," Enderle said. "It's basically Unix with an Apple front end, but from the administrators' point of view, all they see is Unix."

and this is pretty damning too.

Recall that Rob Enderle=Microsoft Apologist

[GillBates0]

Note that Rob Enderle is the author of In Defense Of the Microsoft Monoculture, which was highly debated on /. a couple months back. It surprises me that he should point out the consistency and flexibility of Linux, since his earlier writeup made him look as if he was paid my M$ to mouth major anti-Linux FUD.

Re:Ctrl-C Ctrl-P == Studying?

[Unregistered]

Windows coders are not allowed to get anywhere near linux (or any other code). If a win coder needs to use a linux box, it will have absolutely NO source code on there anywhere. Code contamination scares the hell out of ms becasue if they end up with contaminated code the will be liable for huge amounts in damages and need to rework a large portion of windows(cause everything is intentionally over-integrated) in order to replace the offending code.

History has proven...

[dbCooper0]

Shooting from the hip I'm taking this guess at the model for M$ future:
After WWII, Japan (well, Asia et.al.) learned to imitate US technology - we even helped them! Autos, Electronics, Watches and Cameras are prime examples.
The thing to watch for is what happened next - after Imitation came Innovation...and the popularity of those commodities took a swing to the East. The US economy took a pretty good hit, as I recall!
This is a wake-up call for Linux devs to stay sharp and keep up the steam of progress.

Microsoft is big enough to be a community

[harlows_monkeys]

Microsoft could actually clone the Linux development model, if they wished.

They've got enough programmers that they could simple go "open source" within the company, and that would be a big enough community to get all the benefits cited by ESR in his Cathedral paper.

A few other big companies could also do this if they wished.

Long term, I think, what Linus himself will be remembered for won't be the Linux kernel itself, but for how he managed the project. Hell, I personally know half a dozen people that could have done everything Linus did on the technical side of things, but I doubt any of them would have been the project manager he turned out to be.

Re:Just an organizational change?

[swb]

Perhaps an organizational change is entirely appropriate and the right thing. I don't think that anyone would seriously argue that MS doesn't have any good developers.

What they seem plagued by are marketing-driven technologies that keep getting bolted on to Windows, broadening the code base and making the overall focus of the development harder for anyone to see. This level of integration may make IIS faster or enable easier functionality for some third party development, also makes it hard to define what Windows core is and who's responsible for it.

A group of developers focused on the core of Windows (kernel, networking, filesystem) should be able to better focus on making it work well and keep security at a higher level, among other things.

The real challenge will be who defines what the core of Windows is, and what they define it to be. If they allow the scope of Windows core to be everything you get in C:\ after installing the OS, it won't be more than cosmetic. However, if they define it succintly and at least internally acknowledge that the kernel, the filesystem and the networking code is the core, and other stuff like IIS or Internet Explorer is not, this could mean real benefits for Windows.

Software Design Principles

[killmeplease]

It seems to me that Microsoft is nneding to revamp the way that they make their OS. If you read about the original development team it was 30 or so people for NT 3.1, a managable development team. NT 4.0 was a couple hundred and now the NT/XP team is huge and they are all working on the OS. The bug count last I heard was in the 40,000 - 60,000 region. It is simply to complex a system to keep track of all changes and bugs. In The Million Man Month they discuss there being no silver bullet for creating perfect software, but definitely putting more people on the team will not help.

I think it is great MS is changing the way they make OSs. If MS makes a good OS that is stable and secure, it raises the bar for everyone else to compete. They have to change their method. I think they have there divisions implement changes and Software QA verifies build stability the way that most Application development shops work. This seems impossible on the OS level. Linux is great because they have the UNIX design model to work with, everything laid out in a clean working fashion. The Kernel is so stable because they do not make major changes to the way user space is made because it just has to allow for UNIX applications to run (including X). All of the modules like memory and file systems are layed out. Microsoft has ruined their chance of using this model in Win XP by putting the GUI in kernel space, grabbing the messy registry system from Win 95, and whatever they could hack to make a consumer heavy duty OS.

Perhaps MS could have Linus Torvalds come in and be their OS development consultant for a day. They could elect someone as their Linus to check in every major change. They could add features for 6-12 months at a time, and test and revise changes for the next 6-12 months and blamo they have a Linux like development team.

Re:OK then

[PainKilleR-CE]

It might also be due to Linux Kernel developers not being directly controllable by other areas of development (i.e. projects). For example with the mingling in Win95 of Explorer and the Kernel, for usability and not considering the consequences of such a decision.

Or the fact that the core team for the kernel is quite small and the direction of the kernel is ultimately controlled by this group.

This is likely to be an attempt to minimise the undue influence other departments have over the Windows Kernel development team. This being a good thing as it tries to prevent the projects goals being unduly subverted to make another projects life easier.

They didn't state that this was a kernel development team, though the name implies that the kernel will be part of their responsibilities. The core OS could include quite a bit more than the kernel. Kernel mode alone, in the diagram of the Longhorn OS, includes the kernel, HAL, device drivers, protocols (TCP, IPSEC, etc), portions of the storage and transaction systems, part of the DirectX graphics and audio drivers, input manager, memory, power, config, and process managers, plug and play, LPC, and so on. The 'Base Operating System Services' which includes the kernel mode portions also includes the window manager, GDI/GDI+, Direct3D, the CLR, and more of the storage and transactions subsystems. On top of all of that is the crap that's gotten the most attention recently, including Avalon, Indigo, WinFS, and the network class library (which might also be part of the Base OS services, it's hard to tell in the diagram).

Another thing that could be an explanation of this would be reaction to the antitrust cases not only at the federal level, but also from many of the states. If the Core OS is being developed by a seperate group, they could try to hold this up as an example of isolating the OS and API development from the application development, even within Microsoft itself. Of course, that also could backfire on them with any future efforts along the lines of IE and WMP, because they'd have to put more planning into integrating the needed features into the core OS before slapping together another product.

Some Advice for Microsoft

[Ann Elk]

• Fire Jim Allchin. He has been a liability for years. He wants to turn every project he touches into "Cairo".
• Appoint a competent replacement, preferably not Brian Valentine.
• Do not allow any summer intern "wannabe engineer" code-boys anywhere near the core OS kernel source code without proper supervision.
• Release the core OS kernel as open source. You don't need to release the source for the entire product, just enough to build NTOSKRNL.EXE, NTDLL.DLL, and a generic HAL.DLL. The driver writers of the world will love you for it.

And around it goes again.

[djve]

I remember reading articles back in the 90's (remember, the last decade) that Microsoft was studying Unix and BSD kernels. Some of the impact seems to have shown up but overall it seems like not too much has changed.

Until Microsoft isolates the the kernel calls from user based calls don't hold your breath waiting for a change. Given the legacy code they support I don't think any big improvement will be soon.

Re:Shame/fame is also important...

[psavo]

Not to mention that in OSS every interested party in the entire world can see where you have been lazy and/or stupid... If that is not a strong motivation for churning out quality code I don't know what is!

Yup. When I ported amd76x_pm driver from 2.4 to 2.6 and posted it onto lkml, I got some remarks on overall code quality. All I did was just a quick hack into changed infrastructure so that it would 'just work'. And then I was getting questions like 'what is this', 'why are there large integer constants' etc. I sure learned not to post inadequate code onto lkml then :)

We could have told them if they'd asked...

[El]

Part of their study has been on how Linux has been able to maintain a high level of consistency in the kernel while groups around it maintain maximum flexibility.

Just a guess, but offhand I'd say it through design driven by software developers, versus design driven by marketing and artificial deadlines. In other words, in any organization, the behaviours that get rewarded the most increase, while those that get punished decrease. MS apparently does not reward consistent, flexible design or implementation.