I want to start a discussion under the heading: Should Microsoft start to build a new kernel?

What do you think? If they would do so, old software wouldn't be compatible any more - if not.... well lets see what you think

dean

yes, they should totally scrap "NT", its good but well, a new kernel would be so much better

I sell them the Phinx Kernel. It uses less than 1 MB Ram and 1 MB HDD space. And it can do everything the NT kernel can do.

time for some real New Technology (NT)
NT is becoming Old Technology (OT)

No. I don't think they should ditch the NT Kernel. Optimising the current kernel would suffice. Losing all that native compatibility with a new kernel would end the biggest advantage it has over it's rivals: compatibility.

I think they should optimize the NT kernel. I think they should remove a lot of legacy stuff that's really not needed any more. I think MinWin would have been a step in the right direction, but the last I heard, it was scrapped.

Can't they develop parallely?
have Windows 7, 8 etc. based on the NT kernel
and then Windows Superior (or something like that) "when it's done"?

Will lead to the same results in my opinion with NT 3.1 and Win 3.1 that they'll use it after a few ears just.. but should be better than just stopping and redoing everything.

Microsoft has to be up-to-date with his Customers OS (Windows NTs) but they have to develop something from the scratch

Well MinWin is actually in vista and in Server 2008.

Are you sure? I thought it was a complete rewrite.

MinWin and Server Core are now two completely different kernels. Originally most people called Server Core MinWin. Now its different, MinWin is just a side project like Singularity.

It'd be interesting actually to see MS take Singularity further...

Isn't Singularity a Real Time OS? It's great for research purposes, but as a replacement for NT, it isn't useful. Wasn't Microsoft working on something called 'Sky', or similar, to replace NT?

I think they need to fork the FreeBSD kernel and work with that, adding on all their Microsoft style crap. They also should write a different windowing system (perhaps a GDI/WPF based WS) because X11 sucks.

That is a nice dream, but its never gonna happen. It would be bad for both FreeBSD and Microsoft. For BSD since everyone will be trying to break the kernel somehow. For Microsoft because it would mean probably losing a significant amount of money because of licensing.

Working on the FreeBSD kernel wouldn't be good i think. Imagine:
Linux, Windows and Mac OS X uses the unix kernel. OK there wouldn't be any compatibility problems any more - but there wouldn't also be competition - and then products can't develop any more

Linux IS a kernel, it does not use a unix kernel, and it is not 100% compatible with unix anyway. MacOS is based on a unix-like kernel that has been developed apart from unix for about 7 or 8 years, now.

Linux is THE WAY FORWARD. Didn't someone say that the Windows line is ending in a few years? Linux'll take over! Computers'll be useful again!

I think a rewrite of the kernel is overdue. Some improvements were made in vista, but only added more issues.

just my $.02

Maybe, but if they optimised it properly, and got rid of drivers for devices that weren't needed - maybe replaced outdated sets of drivers with generic ones, if they'd work. It could be streamlined a hell of a lot if they optimised it properly.

Functionally, it's a sound kernel, but they have moved more and more into the kernel - stuff that, if it was in userspace, would enable the kernel to run a lot more eficciently. If device drivers were not in kernelspace, the system would be a lot more stable. Yes, I'm talking about a microkernel. It's a good design, but it's very badly implemented.


EDIT: You think the Windows NT kernel is old? Granted, it's been in development since 1993, but what about Linux? That's been in development since 1991, and they've never rewritten that from scratch? The BSD variants are decended from the same code-pool that spawned Research Unix 1, way back in 1971.

Like I said - good design, but badly implemented. It's probably written in QBASIC or .NET or some tripe like that.

EDIT2: Spelling.

I'm interested in one thing, what would a new kernel add?
What new kernel features do you want? Why do you assume, given that nobody knows how it works, that the way Microsoft has written its kernel is bad?

For example, I like how in the Windows kernel I can reserve I/O bandwidth... Or how it has interupt priority levels. Lastly, I prefer the monolythic kernel approach beacuse it has the potential to be more secure by maintaing more processes as kernel mode. Secondly, there is no need to switch tasks. Thus, monolythics kernels have the potential to be more optomizable.

From the perspective a programer, I wonder what they could actually do. Most of the problems of Windows come from improperly written programs or drivers. To that extent I like the direction of the .Net framework, especially C#, because it recompiles the programs...

Ultimately, what do you expect from a kernel rewrite?

New kernel is what must-do for Microsoft.. .., but not now... I believe there should be parallel work~~~

anyway~ what we hope is having a new kernel and release the source codes for NTs...

I think that they should ditch it. It would solve a lot problems. Only if they kept compatibility with most software.

I don't think they will restart it from 0'

IF yes , it will takes very long time!

Actually, Windows 32 runs on some other operating system, be it DOS or VMSoid. There are virtually no third-party progs that run on the NT kernel (O&O Bluecon is an exception here!).

The original intent of Windows NT was to replace OS/2, and its version numbers were numbered to continue from IBM's OS/2. Since at the time, Windows 3.1 was the DOS based Windows, so it became too the Windows NT version. But vers 1.0, 2.0 correspond to OS/2.

One should recall that most proggies run in Win32, which is an application that runs on top of DOS or WNT, so it should not be possible to replace the kernal.

One notes that the idea of replacing a kernel under a GUI has been mooted elsewhere. Apple's "Blackbird" project was to run Apple's GUI on the NT kernel, OS/X runs on a BSD kernel, Likewise, there is assorted discussion of running OS/2 on a linux kernel or on the NT kernel.

So it's not hard to modularise the kernel: the thing is _already_ modular. What it isn't is a general purpose environment for running anything other than windows, in the same way that MS-DOS 7 is a kernel that can run non-Windows DOS programs.

This guy does a great job explaining the entire thing. Should you be complaining about the Windows kernel without actually knowing anything about it?

Besides, do you want more broken, crashing, and slow device drivers? Then everyone is just going to complain about it like they did Vista.

READ THIS: http://weblog.infoworld.com/enterprised ... ntwin.html

An interesting NT tidbit that I ran into recently, Wikipedia does a better job at explaining it than I do: http://en.wikipedia.org/wiki/Windows_NT ... esignation

Same here but I doubt they will. It was mainly just for research. For anyone who what to know about it, Channel9 (Microsoft's video site) posted a video talking to some of the devs.
http://channel9.msdn.com/shows/Going+De ... tten-in-C/