OSNews

I have strong hopes for Windows 7 (Much the same way that I had strong hopes for Windows Longhorn).

Firstly, relegating backwards-compatibility to a VM (I assume, a rootless, well-integrated VM) would be great. In so many ways, not least because many of the problems with Windows in the past have beed caused by the desire to maintain backwards compaitbility.

Secondly, I hear lots about stripping-back and performance. I'm a potential fan of .NET and WPF (tho I won't go as far as supporting Silverlight), as long as Microsoft can cut away the cruft, and strip down the technology to be lean-and-mean enough to run properly.

Cutting away legacy/outdated code from the kernel/base-userland arena will also allow MS to produce better code, and faster.

If they do what they are promising, and do it correctly, I would like to hope that people won't have to wait for Windows 7 SP1, or SP3 before they get a stable OS.

Lastly, the idea of seperating the GUI/WM support from the OS in general is great. Lots of scope for innovation, while retaining legacy skins for people who can't change.

[/fairyland]

Myself, I seriously doubt this will happen (except in edge cases like removing all user interface, like with Server Core). But knowing Microsoft, I imagine they will spend some man-years (and a few million) making it possible, run it, take one look at the sharp drops in speed and increases in memory usage with the GUI running in the userspace, and quickly ax any thoughts running in that direction.

"But knowing Microsoft, I imagine they will spend some man-years (and a few million) making it possible, run it, take one look at the sharp drops in speed and increases in memory usage with the GUI running in the userspace, and quickly ax any thoughts running in that direction."

Uh, the GUI in Vista runs in userspace, if it is running Aero. MS moved the Graphics drivers back into userspace.

But will you live that long to see them confirmed or not? Because if it takes 5 years for a new Windows version to make it from vaporware to actual product, and a couple more for a couple of SP's... it adds up, man.
Will you still have those hopes 5-6 years from now?

Corey - are you expecting SP1 for Vista to fix it?

So, what features have been removed from Vista, exactly?

WinFS.

Yeah, that's about it, and some less impressive stuff here and there.

But hey, who cares about the truth these days?

Big promised features were yanked. Nothing untruthful about that. What're you on about? WinFS is a big feature to drop, considering how much they hyped it. Also there's the whole Palladium security thing they were going on about. There's a feature we can all be glad about them missing at least. And a lot of features that are just plain useless, like readyboost. (The fact that their performance is so terrible they invented a half-assed system to use removable flash media as a cache is mind boggling. No significant number of people are going to be leaving a usb key or flash card in their computer just to act as cache, especially on laptops, which are the majority these days)

For the most part, the OS was just crap, while they promised the world. Overhyped and underdelivered makes people angry, and rightly so.

Edited 2007-12-14 15:39

WinFS? Avalon? Magic Platform-stability by removing 3rd party drivers from the kernel?

All those OSX-killer features that were promised (to stop people turning to Apple) then were delayed, then cancelled, or just folded into Aero.

There is a reason they started calling it "shorthorn" instead of "longhorn".

Considering Avalon was the codename for WPF I don't know how much credibility you have.

How is there a loss of stability from removing 3rd party drivers from the kernel? That doesn't even make sense.

So you remove drivers written by sometimes unreliable vendors from the kernel and somehow you're going to become less stable? You realize that user land drivers have less potential to ruin the system right?

You realize that kernel mode drivers have a potential to wreak unchecked havok upon a system right?

Moving drivers to user land increases stability.

Then you make the connection to OSX which just seals your fate as a fanboy. Let's look at the facts man.

Oooh, I got a bite.

Considering Avalon was the codename for WPF I don't know how much credibility you have.

Avalon was sold as actually being used in Windows Vista, not just a beta library pre-installed and left for 3rd parties to develop with.

How is there a loss of stability from removing 3rd party drivers from the kernel? That doesn't even make sense.

What I said, was that Microsoft (and it's shills) were claiming that 80% platform problems were cause by 3rd party code running in the kernel space. Longhorn was supposed to have no in-kernel drivers so would have 20% of the stability issues of XP. This has been proved not to be true.

Then you make the connection to OSX which just seals your fate as a fanboy. Let's look at the facts man

I am many things, but an Apple fanboi, I am not. If you must pidgeonhole me, then, I am a Windows XP, Ubuntu, Gnome, E17, Valve Fanboi. The OSX comment was related to the 'coincidental' timing of various software releases, and press releases.

1. Next-Generation Secure Computing Base aka pallladium

2. UEFI support

3. SecureID support

4. PC-to-PC Sync

5. XPS support

PC to PC sync is still there. XPS support isnt there because adobe threatened litigation. UEFI is in SP1. You are right about parts of palladium (elements have been implemented, others havent), and native SecurID support.

Personally I just don't know HOW I'M GOING TO COPE without those HUGE, MUST HAVE features.

Enter the VMs, if done properly, MS will be able to write an essentially new OS, while maintaining, at least passable backwards compatibility.

Linux+vmware...

Had they designed it well in the first place, they wouldn't need to ditch it and start again (which they already partially did with the transition from dos/win9x to nt)...
Apple did the same thing, OS9 and it's predecessors were unmaintainable garbage.
But look at unix, a sensible and simple basic design that's stuck around through many different implementations. Many programs written for those old unixes will still compile and run on modern systems.

Virtualization is the best and really only answer. But I agree, the only way to save Windows is to reinvent it.

It depends on what you mean by "save"... The structure of the Windows ecosystem is inherently flawed. The only way it can work is if change is kept to an absolute minimum. Hardware, UI, and software development paradigms all must remain more-or-less the same for the train to keep rolling down the tracks.

Change is inevitable, and therefore I don't believe there's a way to "save" Windows in the long-run. However, there's definitely a way to keep the gravy train going for at least the next 15-20 years. The key is to resist all the desperate pleas for revolutionary (or even significantly evolutionary) change from their lead-user communities and keep repackaging the same old Windows every 3-5 years.

Keep the change to an absolute minimum. Use forced obsolescence along with carrot-and-stick marketing to move the userbase along a series of minimally-disruptive upgrade cycles. Embrace the fact that compatibility and familiarity--not innovation--is the profit engine of the Windows ecosystem.

There's a fierce loyalty that goes along with brand continuity. It's like those Americans that vote Republican because their family has been voting Republican for five generations. Never mind that they used to be the urban liberals while the Democrats were the rural conservatives. They identify with the brand, so they'll go along with the politics.

Just avoid the urge to change, keep Windows looking and feeling and working like Windows, and Microsoft will do just fine for the next couple decades. Sure, they'll lose ground all over the place, particularly on post-PC computing devices and among tech-savvy consumers, but they'll stay firmly in the black.

Microsoft should be thinking less about how to save Windows in 3-5 years and more about what they're going to do in 15-20 years. That's the timeframe for reinventing Windows. If they keep that target in sight, they'll have a reasonable opportunity to begin a new 30-year Reich.

I guess its time to invoke Godwin's Law and stop talking about this.

Mate, did you actually read the article? it spoke of virtualisation, the use of a virtual machine to host a 'classic Windows session' to allow backwards compatibility.

They're not going to throw out the whole thing and restart, but what they will do is throw out large sections that should have been killed off and removed years ago. Look through Windows at the amount of 'left overs' from NT 4/3 days, heck, even further back than that!

The problem with Windows isn't the operating system itself but the compromises made for compatibility; they know what needs to be removed, they know this all can be done without needing to do major re-writes, the problem is that for the last 10 years they've been putting backwards compatibility ahead of fixing things properly, and doing it right the first time.

Its the old story of 'no pain, no gain' - the current process is the equivalent of a person with a bad back having constant acupuncher and massages when it can be easily corrected with surgery; 2 weeks off work for permanently fixing the problem rather than band-aid solutions.

Loyalty? Come on.....I like Apple, and all, but let's be real. The only reason Jobs was able to get up there and spout out some "Change, or suffer the consequences" mantra had nothing to do with user loyalty. The man would piss on his user base if it would be profitable for him.

No, it had to do with pure numbers. Fact is, the number of developers creating apps for the Mac are orders of magnitude less than those that create apps for Wintel. Example: When was the last time you walked into a manufacturing facility and saw ANY equipment running off a Mac? I'll give you a hint....never. Linux will take that space over WAY before Apple ever will.

I'll agree that in a perfect world, Microsoft should go..."Here it is, change your apps or see ya later." But the fact of the matter is that will never happen. Virtualization is the best shot MS has at breaking away from the backwards compatibility catch-22. in that, you're either compatible and bloated; or your secure, and high performance. Choose one.

Trying to figure out company development from a (deleted!) forum post you once read is a dangerous game

Welcome to the real world. The first to press with information (however factless) about a forthcoming well-known product gets all the attention (and ad-revenue)

"A marketing coup to hide the ripping off of public domain code?"

How do you rip off public domain code?

You claim it as your own. Like as in plagiarism.

I used that term not to refer to the literal public domain, but to refer to things like BSD, MIT code, etc., and other sources that are traditionally treated as if they were public domain... without creating a huge confusing phrase to say the same thing. ^_^;

so you think apple took code covered by bsd & mit licenses an claimed it as its own? would you care to provide any proof? or is this just the same old boring "the bsd-license should be as gpl"-argument?

Yeah, but they shouldn't be to eager to show mockups. They have a long 5 years or so before them, they'll need to spread the mockups pretty thin to last them.

There is nothing wrong with "ripping off public domain code". If it was about BSD code in OSX, I never heard anyone from BSD camp complaining. The license explicitly allows that. Only a fool would implement from scratch something that is available with no strings attached. Only a huge idiot would insist on being original at all costs.

Microsoft could have done much better job if they followed that path in the past.

25Mb is still huge, win3.11 wasn't that big, modern linuxes can be much smaller than that too.

And linux already has a single codebase for disparate projects, the same kernel can be compiled (obviously with different options) for anything from a phone, to an ibm mainframe. That's what makes embedded linux more useful, the ability to easily recompile arbitrary apps.

http://www.rediculous.co.uk/statistics.htm

Was wondering when the spelling nazi would show up!

That strategy kinda ignores the market of people who wrote some mission-critical application in 2001 (or, 1993) and don't want to go through the trouble and probable breakage that rewriting everything would cause.

Of course, you could argue they're probably running Windows 98 (or 3.1) anyway, and just plain aren't in the market...

Exactly,
And its doubtful that these apps would be able to benefit from an operating system upgrade, and take advantage of whatever new features that you would like to use without an update anyways.

That works OK for servers, but what about in-house desktop apps?
Should every desktop in a business be stuck with Windows 98 because one critical app won't run?

Although I do agree that much of Windows is in desperate need of a clean rewrite, back-compat is still important. A good VM solution could work, but it would have to be even better than what Apple pulled off in order to be satisfactory for most users.

Edited 2007-12-14 17:02

That works OK for servers, but what about in-house desktop apps?
Should every desktop in a business be stuck with Windows 98 because one critical app won't run?

Yes, yes it should. The rollout across a company is dependant on all applications working 100% correctly.
If one application does not work correctly the rollout should be stopped and the system in use needs to stay in use.

Edited 2007-12-14 18:44

I don't think that it is about compatibility. Nothing that noble. I think that they are trying to squeeze as much as they can from initial investment in WindowsNT technology.

They need backwards compatibility, windows being the only platform that can run people's proprietary apps is the biggest thing keeping microsoft alive. If users were forced to migrate anyway, then what's to stop them migrating to cheaper alternatives?

Okay it eats up 40MB RAM but cannot show a gfx boot logo... Oh dear!

Well. look e.g. to QNX or MorphOS you'll see what you can achieve with minimal resources.

or to win nt 4. min 16mb ram, 32mb ram advised.
this really isn't a nt-kernel slimmed down to the bare minimum.

Well, it's nice to know they're trying...

(Yes, I've seen a few linux-on-a-floppy things... and a friend once gave me a 13-floppy Windows 95 installer)

"""

"""

40MB was the number thrown around in the Eric Traut talk. But if you watch carefully, it is shown (but not emphasized) that minwin, barely capable of serving a few small, static html pages... and with its ascii art boot logo, was actually running in 76MB of virtual memory, and consuming 61MB of it. Eric, himself, said that he was disappointed by its resource consumption. Nonetheless, it represents the best that MS has been able to manage in streamlining the Windows kernel.

http://www4.osnews.com/permalink?280016t

Edited 2007-12-14 20:55

That's actually kind of sad if true. With all the cash they have behind them they couldn't get performance than that?

So basically they've thrown in the towel as if to say, "I give up, its impossible! we can't slim it down any more!" - they need to go back, and go through the code with a fine tooth comb; when you have operating systems written by far smaller companies with lower foot prints, it speaks of incompetence by those at Microsoft rather than a lack of resources.

We talked about this a while ago. MinWin is NOT an attempt to make a streamlined NT kernel. It is simply a partitioning of the Windows source tree and binaries for organizing internal engineering efforts.

Like much of the news about Microsoft on the Net, this article reflects very little what's going on inside the company.

"""

"""

Yes, we have. But Eric Traut, Director of VM and Kernel Development at Microsoft, who leads the 200 engineer team at Microsoft in charge of kernel development , and MinWin in particular, doesn't seem to be aware of it. In his talk, he says nothing of that, and instead focuses upon the streamlined nature of MinWin. In the talk linked in the story which I linked in my previous post, he repeatedly focuses on MinWin's "small" size and "low" resource consumption. At exactly 49:00, he even claims that the memory stats displayed, and which I have quoted, constitute "proof that there is a pretty nice little core inside of Windows".

He mentions nothing of it not being an attempt at a streamlined kernel, and plenty to imply that it is such an attempt. And he says nothing about MinWin being "simply a partitioning of the Windows source tree and binaries for organizing internal engineering efforts."

"""

"""

With all due respect, what are your sources, which you feel have greater credibility than the man in charge of the team developing MinWin at Microsoft?

Edited 2007-12-15 13:38

I work with some of the people who produce MinWin, I've read many of the specs behind MinWin, and I spend pretty much my entire time in the MinWin source tree.

You're interpreting Eric Traut's words in a way he didn't intend them. I've met Eric several times: he's a very friendly guy and he's extremely knowledgeable about the nitty gritty aspects of OSes and computer architecture (he's one of the primary authors of VirtualPC), but he's not a marketing person and he's not a PR person. He's an engineer, and he was giving an engineering talk to a university audience. People are interpreting his talk as some sort of official marketing, but that's grasping at straws.

Just to be clear... MinWin is not a microkernel and MinWin is not an attempt to strip down the current NT kernel. If you want to find out everything about it before Win7 goes into beta, you'll probably have to change jobs.

Don't know about the ignorance, but I bet that Vista is slower than 3.11 because it was designed to perfom a metric ass-ton more stuff than 3.11. Features aren't free, everything comes at a price. And BTW, there was a time when some people kept using DOS because Windows 1.0 was too heavy for their systems.

I think you'd lose that bet. It's slower just because it is. More features? When the cost of those features is performance, who benefits? Certainly not me, the consumer who foots the monetary "cost". Finally, as has been predicted for years, consumers are voting with their wallets and that's why Vista is the mega flop that it is. All you windoze fanboys need to learn how to use Google and altavista and quit demanding proof that you aren't going to believe anyway. The data is out there. Now, go find it.

Mega flop? 100 million sold is a mega flop?

You realize Vista in one week outsold the entire user base of OSX/Linux correct?

Vista is a lot of things, but it's far from a flop.

Vista boxed set sales were way below expectations. The only reason it got good numbers is because virtually all new PCs come with it pre-installed...well, that is before people demanded to PCs with XP installed instead, which some OEMs did.

That in itself is unheard of in the short history of OSes as a consumer product. "Mega flop" might be an exaggeration, but there's no question it constitutes a flop for anyone except PR reps and fanboys.

I moved from 10.2 to 10.3 to 10.4 on PowerPC systems.

I moved from 10.4.10 to 10.5 on an Intel system

In all these transitions on my Mac, I never experienced the massive slow down which people are experiencing when they go from Windows XP to Windows Vista.

You can't honestly tell me that the changes in Windows Vista are so great as to justify the quadrupling of memory requirements and even more in terms of CPU speed, just to get a decent level of performance?

Take Linux for example, I moved from Fedora 6 -> 7 -> 8, and with each movement forward it either has stayed static with performance on the same machine or improved. Latest statistics in KDE 4.0 has shown a possible 40% decrease in resource usage - even with more eye candy!

I look around the IT market and Microsoft seem to be doing what Joel says; developing for tomorrow, but even tomorrow, the hardware isn't powerful enough to run the bloat. The sad part, Core 2 and AMD make some damn powerful machines; is Microsoft expecting me to believe that a 2Ghz Core 2 or AMD64 4400 isn't powerful enough to run Windows Vista?

Heck, I can pick up a Mac with 1gig and it runs Leopard beautifully - the move to 2gigs by me, rather than it being a must to run at a decent speed, was a choice of mine. I voluntarily upgraded the memory rather than forced by virtue of a software vendor writing loose code.

"I never experienced the massive slow down which people are experiencing when they go from Windows XP to Windows Vista"

Huh, and neither did I - when I went from Windows XP to Vista.

PPC Leopard is slower than Tiger. 32-bit PPC/x86 Leopard is slower than Tiger.

http://arstechnica.com/journals/apple.ars/2007/10/30/benchmark-leop...
OS 10.0 was so slow it was actually painful to use. 10.1 was better, 10.2 faster but not as significant, etc. It was only when Mac got to 10.4 that they had something decent. 10.5 is slower than 10.4 except when it relates to 64-bit which is hardly surpising since 10.4 was not at all optimised for 64-bit.

The sad part, Core 2 and AMD make some damn powerful machines; is Microsoft expecting me to believe that a 2Ghz Core 2 or AMD64 4400 isn't powerful enough to run Windows Vista?

Vista runs very well on and AMD 3200 with 1 Gig RAM and an nvidia 5200. Especially if you install the performance and compatibility update.

Don't know about the ignorance, but I bet that Vista is slower than 3.11 because it was designed to perfom a metric ass-ton more stuff than 3.11.

Yeah, great point. I totally agree.

Features aren't free, everything comes at a price.

Some consider performance or small memory footprint a feature.

And BTW, there was a time when some people kept using DOS because Windows 1.0 was too heavy for their systems.

I guess the cycle of bloat continues, then.

"AMD 286-20 with 768MB of ram"

Ummmms. Are you sure it was AMD?

Also - where did you find such 268 system with 768 MB of RAM? I wonder if it supported even 16 MB at all.