Microsoft Linux - it's official. | The Coffeehouse

Rossj

Sorta. Well okay, not really. But Microsoft *is* making deals with Novell to allow integration* between Linux and Windows.

This is a good thing. Right?

* It's actually not really very clear. But something like that.

SecretSoftware
Code to live, but Live to code.

If Windows had the kernell of Linux, Windows would become more stable. I want an OS with Linux stability, but the UI of Windows Big Smile

Simplexity in design.

Cornelius Ellsonpeter

So is this some sort of virtual machine type setup? Windows as the platform...Linux software running on top?

CannotResolveSymbol
{insert caption here}

Just wait until 2pm PST and all will be revealed...

(Reuters):

Microsoft plans to make a major announcement later on Thursday, but declined to comment on the article.

(Presspass):

Media Alert: Microsoft CEO Steve Ballmer to Deliver Industry Announcement Today Details of the announcement will be provided during a press conference.

SAN FRANCISCO — Nov. 2, 2006

What:

Microsoft Corp. CEO Steve Ballmer will deliver an announcement during a press conference today in San Francisco.

When:

2 p.m. PST, Thursday, Nov. 2, 2006

Webcast:

A live webcast will be available. After the conference has ended, the video will be available for streaming at the same link for at least six months.

Rossj

What I'd love to see, but I know won't happen is Microsoft to consider letting the Vista Kernel guys talk openly (with code) with the Linux guys. Here's why...

I don't think the Linux guys come across as arrogant enough to think that there isn't anything they could learn from access to the Vista kernel, and I think the reverse *might* also be true. I'd *love* to see a Microsoft working together on the kernel with the OSS community and being able to expend more effort on the desktop environment and less on the kernel level stuff - although paying people to work on an OSS kernel would be good.

At some point I believe (and have no proof) that Microsoft will have to give up any pretension of full back compat - supporting Windows 3.1 code in the version after Vista makes little sense (at least to me). Do you see the OSX community complaining that they can no longer run classic apps? Sure the market share is smaller but there was a path through the pain and people accepted having to run some apps in the classic environment - and now they just don't seem to care.

Microsoft collaborating, or paying MS employees to work on the Linux kernel benefits the Linux guys, it benefits Microsoft and it benefits the end user (where the differentiator is the desktop environment - an area where Microsoft is likely to win).

Of course I am dreaming, it'll never happen.

PaoloM
Hypermediocrity

But the things is, what would be the benefit of using the Linux kernel instead of the NT one?

Let's not even think about backwards compatibility, where is the Linux kernel better than NT?

Rory
Free Tibet While Supplies Last

SecretSoftware wrote:
If Windows had the kernell of Linux, Windows would become more stable. I want an OS with Linux stability, but the UI of Windows Simplexity in design.

I disagree. Strongly.

Linux is cruft upon cruft upon cruft.

I wish we could, in the entire tech world (Apple, MS, OSS, etc.), throw everything away each year and start over with what we've learned.

BeOS was a cool example of a very modern OS (in some ways). I wish we all had the convenience of getting to start fresh.

Using some old crusty kernel - like Linux - isn't my idea of a good move. And I don't mean to bash Linux, but when I used to hang out with my kernel hacker buddies, the one thing they told me was that they loved OSS, they loved the ideas, but that the software was a mess.

Nothing wrong in my mind with Unix, but Linux? No thanks.

Again, though, if we could just swap kernels, I'd rather see a kernel created from scratch based upon modern ideas, modern code, and not the leftovers of ideas and structures created last century.

The problem is that it turns out that it kind of angers customers when you tell them they're going to have to throw away their old software.

So this is just a dream...

staceyw
Before C# there was darkness...

"If Windows had the kernell of Linux, Windows would become more stable. I want an OS with Linux stability, but the UI of Windows

Simplexity in design."

Wrong argument. The NT kernel has never been the issue. In fact, the kernel is sweet and has a much better design then the monolithic linux kernel (i.e. spegetti). The perceived issues with Windows have always been driver or application (i.e. user mode stuff) related and all the UI goo that runs ontop of the kernel. If you ran NT headless or with just a console as the shell it would never crash. I had NT, 2000, and 2003 servers that never crashed when I managed them because I treated them as production machines and not as a shareware test bed.

staceyw
Before C# there was darkness...

Sounds like Novell may need another cash injection...

BryanF

Rossj wrote:
What I'd love to see, but I know won't happen is Microsoft to consider letting the Vista Kernel guys talk openly (with code) with the Linux guys. Here's why...

I don't think the Linux guys come across as arrogant enough to think that there isn't anything they could learn from access to the Vista kernel, and I think the reverse *might* also be true. I'd *love* to see a Microsoft working together on the kernel with the OSS community and being able to expend more effort on the desktop environment and less on the kernel level stuff - although paying people to work on an OSS kernel would be good.

At some point I believe (and have no proof) that Microsoft will have to give up any pretension of full back compat - supporting Windows 3.1 code in the version after Vista makes little sense (at least to me). Do you see the OSX community complaining that they can no longer run classic apps? Sure the market share is smaller but there was a path through the pain and people accepted having to run some apps in the classic environment - and now they just don't seem to care.

Microsoft collaborating, or paying MS employees to work on the Linux kernel benefits the Linux guys, it benefits Microsoft and it benefits the end user (where the differentiator is the desktop environment - an area where Microsoft is likely to win).

Of course I am dreaming, it'll never happen.

I'm all for more industry collaboration, but I don't see why Microsoft would make an effort to port Windows onto the Linux kernel. Even if we assume there's no technical reason it couldn't work, I'd suspect the potential payoff would be so marginal that it really wouldn't be worth the engineering effort. While NT is more complex, that complexity isn't necessarily all backwards compat cruft; Windows's driver model is much more flexible, for instance. Also, as important as Windows is to Microsoft, I think they still benefit considerably from having full control of the evolution of the kernel. It's a competitive advantage that enables them to do things that, in the short term, lead to features such as kernel-level transactions, and, in the long term, perhaps more towards Singularity.

(BTW, I believe 16-bit compat actually has been removed from 64-bit versions of Windows. My understanding is that the Virtual 86 mode needed to run those older apps isn't available to x64 processors in 64-bit mode.)

Detroit Muscle
Wii!

PaoloM wrote:
where is the Linux kernel better than NT?

http://en.wikipedia.org/wiki/Symmetric_multiprocessing

JasonOlson

Rossj wrote:

Microsoft collaborating, or paying MS employees to work on the Linux kernel benefits the Linux guys, it benefits Microsoft and it benefits the end user (where the differentiator is the desktop environment - an area where Microsoft is likely to win).

Unfortunately, I'd imagine that there are real-world legal ramifications that would prevent this from happening (especially around the definition of what constitutes a "derived work") Sad

.

JasonOlson

Detroit Muscle wrote:

PaoloM wrote: where is the Linux kernel better than NT?

http://en.wikipedia.org/wiki/Symmetric_multiprocessing

Uhhhh, if you read the article you yourself pointed to, Windows NT/2000/XP is listed under the "Capable Operating Systems". Am I missing something?

Sven Groot
My name has 9 letters. Coincidence? I think not...

Afaik, Linux doesn't have the technical limitation of Windows that limits it to 32 logical CPUs for the 32 bit version and 64 for the 64 bit versions (x64, ia64). Maybe that's what he meant?

Andrew Davey
www.aboutcode.net

Novell sponsor the Mono project. I would love to see Microsoft helping out the Mono team to get .Net 3.0 fully supported under Mono.
I've only just started experimenting with Mono and so far I am very impressed.
I reckon a little nudge from MS would make writing true rich UI, cross platform applications a reality.

Rossj

PaoloM wrote:
But the things is, what would be the benefit of using the Linux kernel instead of the NT one?

You wouldn't have to pour thousands and thousands of man hours into maintaining it.

You can streamline and focus on the areas that desperately need addressing.

You can leverage resource that you don't own and don't pay.

You can make use of technology in the kernel that *may* be better implemented than in NT.

How about we turn this around - you tell me what the disadvantages are of using the Linux/BSD kernel?

PaoloM wrote:
Let's not even think about backwards compatibility, where is the Linux kernel better than NT?

I don't know, I've never seen any of the NT source.

Avoiding stupidly expensive and crufty back-compat is what I am trying to suggest. I also didn't say the Linux kernel was better than the NT one, but it could be with input from the NT guys, and it would allow Microsoft to concentrate on the desktop (you'll probably disagree but I still think Windows can be kludgy and unintuitive and intrusive*) where it is needed. Bear in mind that I think Microsoft continuing to develop IE7 is pointless (my opinion and I am entitled to it), so why shouldn't I hold the opinion that Microsoft could streamline in other ways by cutting out those things they don't need to develop.

My guess is that Microsoft are going to announce some sort of strategic partnership that'll no doubt cause ruptions in the non-commercial parts of the OSS community. The Novell shareprice jumped a little....

* And no I am not suggesting that none of these are true about OSX, apart from the intrusive thing - which it isn't. It only ever tells me when stuff goes wrong, never when stuff goes right.

BryanF

Sven Groot wrote:
Afaik, Linux doesn't have the technical limitation of Windows that limits it to 32 logical CPUs for the 32 bit version and 64 for the 64 bit versions (x64, ia64). Maybe that's what he meant?

That could be, but I first heard that so long ago I don't know if it's still true off the top of my head and besides I think we've got a few more years before that becomes a major issue. I can't see how it would take less effort to move everything over to Linux than to fix that flaw in NT. Either way, you're going to have to deal with driver breakage, and with Windows, one of the design goals of WDF was to abstract away internal data structures so they could be modified in the future, so if it isn't fixed already, then Microsoft may be opening a window (no pun intended) to do so in the future.

PaoloM
Hypermediocrity

Rossj wrote:
You can leverage resource that you don't own and don't pay.

Ack! WARNING WARNING

The absolute first thing that they taught me at Consulting 101 is:

YOU

DO

NOT

OUTSOURCE

YOUR

CORE

BUSINESS

Smiley

Ever wondered why Apple and Microsoft have all this Not Invented Here syndrome? Precisely for this reason.

As an example, see the difference in IP licensing between the Xbox and the Xbox 360. With the first one, Microsoft had to rush it to get to the market and took dependencies on Intel, nVidia, etc. The second one was more deliberatly thought and they basically own most of the IP rights on it.

Again, if your core business is in selling software (not as a reseller), you build it on your own. Always.

Sven Groot
My name has 9 letters. Coincidence? I think not...

BryanF wrote:

Sven Groot wrote: Afaik, Linux doesn't have the technical limitation of Windows that limits it to 32 logical CPUs for the 32 bit version and 64 for the 64 bit versions (x64, ia64). Maybe that's what he meant?
That could be, but I first heard that so long ago I don't know if it's still true off the top of my head

It's still true. It's because the CPU affinity for threads is set using a word-sized bitmask.

While it isn't an issue for 99.9999999% of usages, it does mean that the *nixes including Linux still have a scalability advantage in the supercomputer arena. I mean, what use is buying an HP Superdome with 128 Itanium2 CPUs if you can't use them all?

Rossj

PaoloM wrote:

Rossj wrote:You can leverage resource that you don't own and don't pay.

Ack! WARNING WARNING

The absolute first thing that they taught me at Consulting 101 is:

YOU

DO

NOT

OUTSOURCE

YOUR

CORE

BUSINESS

I know this. but ..

THE

KERNEL

IS

NOT

YOUR

CORE

BUSINESS!!!!!

If you think it is, then we'll just have to agree to disagree.

PaoloM wrote:
Ever wondered why Apple and Microsoft have all this Not Invented Here syndrome? Precisely for this reason.

Apple took an existing Open Source kernel and modified it whilst keeping it available to the outside world (who just happened to not be interested in working on it). How is that NiH?

PaoloM wrote:
Again, if your core business is in selling software (not as a reseller), you build it on your own. Always.

Always. When it makes sense.

Personally when it is a toss up of telling the client, you can buy a third party library for $500 or I can write it for you which will cost $1000 and take 1 month - which do you think they want? I would never suggest (if they are a software house) that they build every single component that their software uses or depends upon.

Besides - we (the outside world) *have* to rely on you even if we write software, I wouldn't even dream of thinking - right, need to re-invent everything now because I don't want this 6-dev company to have to rely on Microsoft.

BryanF

Sven Groot wrote:
It's still true. It's because the CPU affinity for threads is set using a word-sized bitmask.

While it isn't an issue for 99.9999999% of usages, it does mean that the *nixes including Linux still have a scalability advantage in the supercomputer arena. I mean, what use is buying an HP Superdome with 128 Itanium2 CPUs if you can't use them all?

OK. Thanks for the clarification.

JasonOlson

Sven Groot wrote:

BryanF wrote:

Sven Groot wrote: Afaik, Linux doesn't have the technical limitation of Windows that limits it to 32 logical CPUs for the 32 bit version and 64 for the 64 bit versions (x64, ia64). Maybe that's what he meant?
That could be, but I first heard that so long ago I don't know if it's still true off the top of my head

It's still true. It's because the CPU affinity for threads is set using a word-sized bitmask.

While it isn't an issue for 99.9999999% of usages, it does mean that the *nixes including Linux still have a scalability advantage in the supercomputer arena. I mean, what use is buying an HP Superdome with 128 Itanium2 CPUs if you can't use them all?

My question is what is more popular (I ask because I don't know): SMP or AsyncSMP. IIRC, Beowulf clustering (isn't the computing method Google search uses still?) is ASMP, _not_ SMP. So, this isn't a limitation. And in the ASMP realm, we have Compute Cluster Edition which is starting to pick up steam in the HPC area.

I thought with how cheap computers have become, that people were leaning more towards buying a crap load of smaller PCs and wiring those together via ASMP rather than spending BOATLOADS of money on a 128-way machine in a single box. Isn't this one of the trends that has largely put Cray out of business?

Disclaimer: This is utter speculation since I know JACK SQUAT about the HPC field .

Rossj

Stream is here if anyone is watching.

JasonOlson

Rossj wrote:

PaoloM wrote:

Rossj wrote: You can leverage resource that you don't own and don't pay.

Ack! WARNING WARNING

The absolute first thing that they taught me at Consulting 101 is:

YOU

DO

NOT

OUTSOURCE

YOUR

CORE

BUSINESS

I know this. but ..

THE

KERNEL

IS

NOT

YOUR

CORE

BUSINESS!!!!!

If you think it is, then we'll just have to agree to disagree.

PaoloM wrote: Ever wondered why Apple and Microsoft have all this Not Invented Here syndrome? Precisely for this reason.

Apple took an existing Open Source kernel and modified it whilst keeping it available to the outside world (who just happened to not be interested in working on it). How is that NiH?

PaoloM wrote: Again, if your core business is in selling software (not as a reseller), you build it on your own. Always.

Always. When it makes sense.

Personally when it is a toss up of telling the client, you can buy a third party library for $500 or I can write it for you which will cost $1000 and take 1 month - which do you think they want? I would never suggest (if they are a software house) that they build every single component that their software uses or depends upon.

Besides - we (the outside world) *have* to rely on you even if we write software, I wouldn't even dream of thinking - right, need to re-invent everything now because I don't want this 6-dev company to have to rely on Microsoft.

To me, all this is almost moot. As time progresses, I believe we are starting to see the commodidization (sp?) of operating systems. OSes are going to become less and less "important" in the consumer's eye. The question is, can we (and other companies) adopt fast enough to find other revenue streams.

Rossj

JasonOlson wrote:

Disclaimer: This is utter spe

culation since I know JACK SQUAT about the HPC field .

According to this site, Microsoft has a little way to go to make some headway into the HPC market. I am sure they'll make a dent at some point though.

Operating system Family	Count	Share %	Rmax Sum (GF)	Rpeak Sum (GF)	Processor Sum
Linux	367	73.40 %	1517733	2500463	427171
Windows	2	0.40 %	6196	10512	1560
Unix	96	19.20 %	493113	725733	132768
BSD Based	4	0.80 %	49852	55480	6136
Mixed	26	5.20 %	685932	869652	298704
Mac OS	5	1.00 %	37228	60995	7256
Totals	500	100%	2790054.02	4222834.82	873595