First Details of Windows 7 Emerge

Some small but significant details of the next major release of Windows have emerged via a presentation at the University of Illinois by Microsoft engineer Eric Traut. His presentation focuses on an internal project called "MinWin," designed to optimize the Windows kernel to a minimum footprint, and for which will be the basis for the Windows 7 kernel.

Re:that sounds good but..

[Raineer]

OS X isn't linux based.

Need a smaller knife to split those hairs? I don't see where anyone mentioned it was, regardless. Unix ~= Linux.

Re:Lesson in MS Counting

[Cowclops]

3 = 3 9x = 4 2k/xp = 5 vista = 6 7 = 7 Nuff said.

Re:Lesson in MS Counting

[Anonymous Coward]

Actually, the proper lineage is:
Windows NT 4, Windows 2000 (NT 5), Windows XP (NT 5.1), Vista (NT 6), 'Windows 7' (NT 7)

Re:that sounds good but..

[Space cowboy]

"has mac done this or is it just that the OS on a linux bas system is just plain faster"

The implication that the Mac might have got rid of the BIOS (and hence gained speed) is tied to "a linux-based system is just plain faster". You could easily read that as suggesting the Mac is Linux-based.

FWIW, the Mac doesn't use a BIOS, it uses EFI (Extensible Firmware Interface) these days. And it's not Linux-based either.

Simon.

Re:that sounds good but..

[CSMatt]

Ditching the BIOS will only save a few seconds at best. In my experience most of the time spent waiting for Windows to be usable is not waiting for the drivers to load but waiting for all of those (mostly unnecessary) background processes to start up after logging in. And in any case those few seconds are vital if someone wants to boot from a CD. The American Megatrends BIOS in my built computer literally loads the OS almost immediately after being turned on when the quick test is used instead of a full POST. I found this so annoying that I had to actually enable the full POST along with a custom boot image to slow it down enough to be able to get those precious seconds back, which was not an easy task since I had pretty much a half-second to enter the BIOS setup.

Re:that sounds good but..

[Anonymous Coward]

While the grandparent post is a non-sequitur, it is a true statement. Linux != UNIX != OS X, and that isn't splitting hairs. OS X has a UNIX-like layer on top of its kernel. The kernel, however, has nothing to do with (and is nothing like) Linux. OS X has attributes which resemble Linux, but it is definitely not based on Linux.

Re:Lesson in MS Counting

[hyeh]

Actually there are 2 Windows lines...

MS-DOS Based
1.x, 2.x (Windows/286, Windows/386), 3.x, 4.0 (95), 4.1 (98), 4.9 (Me)

NT Based
3.1, 3.5, 4.0, 5.0 (2000), 5.1 (XP), 6.0 (Vista), 7

Re:microkernel?

[mdmkolbe]

Sorry to disappoint you, but this is not a microkernel. This is just slimming down on what is packaged with the OS.

Think more along the lines of sliming Red Hat Linux down to the size of Damn Small Linux, except right now Windows has the shell, X11, Gnome, etc. all running in kernel mode.

The continuum looks something like this:

1. Everything (shell, graphics system, window manager, etc.) in the kernel. (This is Windows as it is now.)
2. Modularization to the level where different configurations can be distributed. (This is what it sounds like Windows 7 is going for.)
3. Major subsystems (shell, etc.) moved into userspace, but still with a lot of stuff in kernel space (drivers, filesystems, etc.). (This is what Linux is now.)
4. Everything moved into userspace that possibly can (Usually only threads, IPC and address spaces are left in kernel space). (This is a true microkernel.)

Re:Rinse, Repeat

[ozmanjusri]

That's got to be believable!

Here's another one for ya... http://www.roughlydrafted.com/RD/Q4.06/4E2A8848-5738-45B1-A659-AD7473899D7D.html [roughlydrafted.com]

...and here's 364,000 more http://www.google.com.au/search?q=microsoft+vaporware&ie=utf-8&oe=utf-8&aq=t&rls=org.mozilla:en-GB:official&client=firefox-a [google.com.au].

Re:Good intentions

[Kris_J]

Actually, VMs can reduce overhead for IT. For each legacy app, build a VM image and then just deploy that image to any of the PCs that needs it. Much more reliable than trying to run it directly through some dodgy legacy support.

The Microsoft secret to success

[jmorris42]

> So Microsoft tells something about the next version of Windows not long after the people
> have noticed that their current version isn't all that it's made up to be?

Duh. They have been doing this same bait and switch for the life of the company.

Step One. Release wonderous New Version! It is THE must have thing.

Step Two. Everyone realizes it sucks but their money is already in Bill's pocket. And everyone realizes they have no choice but to adopt the new product anyway because of the three year hardware replacement cycle and the illegal (as certified by a US court) bundling agreements with the OEMs that continue to this day. Especially in the case of their OS but to a lesser extent with Office and the other crap they peddle.

Step Three. Microsoft begins hinting about the upcoming new version. It will fix all of the (not quite admitted) problems with current version AND add exciting new must have features. And it is coming Really Soon.

Step Four. Have their minions in the trade press obsess about Upcoming new version. All complaints about Current version are answered with "But Upcoming version will be out soon and will fix that problem." After a year or two make sure to begin writing reviews for competitors products by comparing them to features that Upcoming version will be shipping "Any day now". By this point EVERYONE must be lamenting how crappy the shipping version is to help generate the NEED to upgrade when the new version ships.

Step Five. As the death march to release continues and feaures get cut, spin it as a good thing. (We are focusing on the needs of our customers, blah, blah.) Now that there is beta (anyone else would rate it pre-alpha but.....) code get the drumbeat ramping up in the press with lots of articles and screenshots. Will your hardware be compatible? Can life as you know it continue without the exciting new features? Etc, blah blah.

Step Six. The product finally releases... See Step One.

Re:that sounds good but..

[McFadden]

Linux != UNIX != OS X

Actually, from next Friday... Leopard gets UNIX certification [macworld.com] Mac OS X will be officially Unix.

Re:that sounds good but..

[sumdumass]

I used to have a program that analyzed the boot log file from a windows 98/ME machines and pointed to anything over a set time frame. It diplayed everything in the boot log but it let you filter it to specific times, failures and all that. Often the generic devices that used system memory and processing power took the most times to load. Taking the modem out and replacing the on board stuff with good full blown cards would decrease load times enormously.

It is possible you had a sound card that just wasn't a full blown hardware sound and off loaded a bunch of stuff onto the system's processor and memory.

Of course the different types of boot logs on NT machines didn't work so it cannot look at XPs boot logs. I haven't found anything like it for 2000/XP either. Which really sucks because often the boot log can show all sorts of problem areas that could lead to other glitches in the OS.

Re:ah! just in time

[GaryPatterson]

Mod parent to +11 'insightful reference to history'

It's a common tactic from Microsoft. When there's nothing to say and a competitor may get some PR from a tech media looking for something to write about, come out with something about a product that's on the drawing board, or is only marginally closer to release than the drawing board.

Re:Can I get a little insight, please?

[ozmanjusri]

What is the standard memory usage for a Linux Kernel

Download sizes;
Version 1.0 * Current: 1.0.9, 16-Apr-1994 * Size: 1.3 KB(bz2)
Version 2.6 * Current: 2.6.23, 09-Oct-2007 * Size: 5.8 MB(bz2)
http://www.linuxhq.com/kernel/ [linuxhq.com]

In use;
Linux kernel 2.6.23, 1.8M on disk and 2.3M in RAM.

I don't have a copy of the 1.0 kernel to compare with, sorry.

Re:Lesson in MS Counting

[Lord Ender]

No! The Win9x line is dead. 2000, XP, 2003, and Vista are the Windows NT line.

Re:microkernel?

[Lisandro]

The nice thing about microkernels is that the kernel itself it's completely isolated from the rest of the software running on it - even device drivers, which can (and are) a breaking point in common monolithic kernels. Instead of doing direct calls to the kernel, the software now uses a system of messages.

This, of course, works just fine and it makes the kernel rock solid, but makes system calls slower. I'm guessing that when Win7 is released hardware will be fast enough that this will be a non-issue (hell, it might even not be one now), but the point is, a "regular" kernel will almost always outperform it on the same hardware.

Oops, forgot NT 4.0

[ben there...]

Should be:

3: NT 3.51
4: NT 4.0
5: 2000 (5.1: XP)
6: Vista
7: Win7

Or just look at this [wikipedia.org]. I should have google'd it first. It's all right there.

Re:Lesson in MS Counting

[swordfishBob]

Where's NT 3.1, NT3.5 ?

Re:Wrong family line

[pushing-robot]

Consumer line:

Windows 1.0
Windows 2.0
Windows 3.0
Windows 3.1
Windows 95 (v. 4.0)
Windows 98 (v. 4.1)
Windows ME (v. 4.9)
Line killed off.

Business line:

Windows NT 3.5
Windows NT 4.0
Windows 2000 (v. 5.0)
Windows XP (v. 5.1)
Windows Vista (v. 6)
Windows "7"

There were no NT versions prior to 3.5 because the first NT was released after Windows 3.11, and Microsoft wanted their numbering to be consistent. NT 3.5 coexisted with Windows 3.x (and shared the same GUI design), NT 4.0 coexisted with Windows 4.x, and then MS killed off the "Consumer" Windows line, leaving the NT line to fill versions 5 and 6.

Re:Lesson in MS Counting

[twistah]

Don't forget NT 5.2, which encompassed the ever-popular Windows 2003 Server, as well as several 64-bit releases of XP. NT6.0 also encompasses Windows Server 2008.

Re:that sounds good but..

[InvalidError]

What does a BIOS do? It does POST, lets the user customize some low-level system settings and puts the system in a known state before loading the OS's (boot-)loader.

An x86-style legacy BIOS does the same fundamental things as an x86-style EFI BIOS, the only major differences being the BIOS APIs, how the boot process is structured and the fact that EFI is not backwards-compatible on its own. Other than that, a BIOS, by any other name, is still a BIOS. EFI simply has fewer kludges and ties to legacy x86 hardware.

BTW, a few weeks ago, I read an article about some MoBo manufacturers considering adding 512MB-2GB of flash memory to boot an embedded Linux desktop from the BIOS for disk-less web-browsing and other stuff... a BIOS with embedded Linux does not seem that far-fetched, we only need 1GB firmware hubs to plug into Intel's chipsets and hope we will not need to flash our 1GB BIOS too often.

Re:Rinse, Repeat

[ozmanjusri]

Yep. Business as usual...

Cairo:

1. Announced in 1991 to distract from the lack of anything dramatically new in Windows 3.0.
2. Expected in 1994. Pushed to late 1995, pushed to late 1996, intended to debut in 1997. Changed to a vision.
3. Core features dropped. Ended up as polish on the existing Windows 3.0: Windows 95.

Longhorn:

1. Announced in 2001 to distract from the lack of anything dramatically new in Windows XP.
2. Expected in 2003. Pushed to 2004, 2005, pushed to late 2006, intended to debut in 2007.
3. Core features dropped. Ends up as polish on the existing Windows XP: Windows Vista.

Windows 7:

1. Announced in 2008 to distract from the lack of anything dramatically new in Windows Vista.
2. Expected in 2010. Pushed to 2014, 2015, pushed to late 2016, intended to debut in 2017.
3. Core features dropped. Ends up as polish on the existing Windows Vista: Windows 7.

Fraud as a Business Plan The magic of the Internet is helping to point out the tragic fallacy of believing in Microsoft's promises. Microsoft assures us that it won't ever slip half a decade between operating systems again, but what about the fact that that's all it has ever done?

http://www.roughlydrafted.com/RD/Q4.06/4E2A8848-5738-45B1-A659-AD7473899D7D.html [roughlydrafted.com]

Re:Virtualised Legacy

[Bert64]

There are even some companies with legacy systems tied to Amiga hardware... Using the proprietary Zorro bus. When these machines die, there are no replacements, not even any compatible replacements.

The companies who originally chose these systems, the Amiga based devices, and the customer records database you talked about, made a huge mistake in selecting proprietary technology, and are now paying the price. You'd think enough time has passed for the industry to mature, but people are still choosing proprietary tools with no thought for the risks in the future.

Luckily proprietary hardware is all but dead, and hopefully software will go the same way.

Re:that sounds good but..

[yakumo.unr]

Embedded linux is happening already with some boards like the new asus p5e3 [asus.com]

Re:So what?

[ettlz]

Although there's nothing fundamentally wrong with the Linux Monolithic Kernel (especially since it's behaving more and more like a microkernel these days)

Exporting the odd service or two (like FUSE) does not a microkernel make. And modern NT is not a microkernel by any stretch. Read what AST has to say. If anything, NT's gigantic Executive makes it even more monolithic than Linux. Anyway, the best solutions often borrow ideas from all over the shop.

Linus has admitted that were he to start from scratch, it wouldn't be monolithic.

Citation, please.

Re:Lesson in MS Counting

[Anonymous Coward]

You missed out NT 3.51, NT 5.2 (2003) and the upcoming NT 6.? (2008).

I must be in the sweetspot; everyone else here is too young to remember or so old they've started to forget.

Re:Wrong family line

[TheNetAvenger]

Since the system took 10 minutes to get to the login prompt

You must of had some serious hardware issues. If you had 12mb of RAM, or 16mb of RAM NT 3.1 booted as fast as the DOS Win 3.1, and yes even the server version, as there was even less distinction between the workstation and server versions then.

We moved all our development and tech employees and their respective servers to NT 3.1 in 1993, and trust me this would never of happened if it took 10 minutes to boot.

Average system Specs: 486-33/66 12/16mb RAM...

PS Compared to our Novell Servers, NT file operations (especially remote booting clients) was 2-4x as fast as Novell. Trust me, MS didn't 'dent' the Novell market because NT sucked. Not only was it faster, easier to manage for small business but was a great application server platform, something Novell 'never' got.