OSDev.org

I have never had any problems with microsoft software. I play there games and use windows 98.

dorip,
I have personaly done tests on FAT16 vs. FAT32. FAT32 is slower. Mainly because of the cluster size(if I remember right) the auther of Optimizing Windows for Games Graphics and Multimedia explains this very clearly in his book(in fact, he even has a good chapter on what he thinks MS needs to do to improve Windows). Now I'm not saying that FAT16 is great, it's faster than FAT32 but can't be used on partitions greater that 2GIG, and it is slower than several other file systems. I think that any new OS that wants to fight Windows will need a better file system(but leave FAT12 and floppies alone for compatibility reasons).

Not make a new OS compatible with Windows ???? Most of the software made only runs on Windows(especily games). Besides am I going to want to spend hundreds of dollars(US) on a bunch of new software made to run on some new OS when I could keep what I already have? Of course not!

I would not base a new OS entirerly on Linux. Some things I would probably use(like the TCP/IP stack code from FreeBSD, a Linux based OS. Most think including MS that it is the best ever written), but most I would make most of it myself.

If you do talk about the shortcomings(and plain stupid) stuff about Windows, it can help keep you from making some of the same mistakes MS has made(like the "Start button", I thought I already had my computer started ).

K.J.

dorip,
Not make a new OS compatible with Windows ???? Most of the software made only runs on Windows(especily games). Besides am I going to want to spend hundreds of dollars(US) on a bunch of new software made to run on some new OS when I could keep what I already have? Of course not!

U know, tho, there are at least 10 projects I've seen trying to do what you say is impossible.
I quite agree about buying new software, Win software already costs $$$, so why buy more?

GAMES are the only thing that deserve to run on the new os. Games can be moved from one folder to another and stil run fine(no dlls in the system folder...). I think that normaly we should provide office software. Programs like autocad or corel draw: we have a few oprtions here:
1
Talk to the companies and get them to recompile it and link it with your linker, or get them to let you do it and then give the cd with the program to anyone that buys the os and brings in a valid certificate of autenticity(this is good marketing as well). We also let the company(crorel) to sell it to people that didn't have it.
2
Make a program that would do relocations and stuff in the executable and dlls during the instalation of the program(this would be a pain in the @$$)
3
Make a patch and distribute it on the internet(stil bad as you still use bad parts of windows system)

The best one is the first one.
Drivers:
geta all the hardware companies to recompile their windows drivers with your compiler and linker(easy).

1. Um, no.  People won't recompile for hyped, hyped Linux when it's technically possible, why would they for a no name OS?  Have you ever tried recompiling the source for something for an OS that isn't explicitly supported?  Here's news: It's NOT easy.  In fact, it can consume MONTHS of work, no matter how compatible the two are.

2. This is akin to virtualization of the CPU.  You don't actually have to do symbol relocation in nearly all cases if your kernel simply supports loading different binary executable formats and the call mechanism of the OS you are trying to clone (Windows).  Of course, then you have to have libraries that support all those calls, which either means that you reverse engineer the API and build your own libraries (like WINE), or expect the user to supply the windows ones, which would introduce the retail cost of windows to the system.

3.  This also doesn't make any sense.  You'd have to make patches for each prog.  Why bother?  Just make your kernel support PE and the windows calling mechnism (Pascal-like, not C-like).

geta all the hardware companies to recompile their windows drivers with your compiler and linker(easy).


Slow down, cowboy.  Have you ever written a driver?  Have you ever even looked at a driver API?  You have NO IDEA how what you just described and how it is definitely not 'easy'.  Are you proposing building a kernel that supports both a WDM driver model AND yours?  Or just WDM?  Are you talking about building a 'Freedows' ?  Drivers are bound to their OS.  You generally think of it as half code to speak to the hardware and half code to communicate to the OS.  Provided that you actually managed to support the WDM driver model without completely crippling your kernel, it's REALLY unlikely that you'll be able to talk anyone into supporting your OS.  Hell... Nvidia won't let source code that supports their products to leave their building (literally) much less allow you or anyone else enough access to recompile their driver.

Here's what I suggest: Build something and come back and thensolicit help.  I can't see how anyone with the expertise to build something like an OS would spend the time and (massive) effort to follow your lead until you show them that you have more than just grand visions.

I think that Bart Grantham is right. No hardware manufactor would give out their driver code for you to compile to get it to work on your OS. Espeicily graphics/video card manufactors(just look at alt.os.development, when people want to know how to take control of a NVidia or ATI card the answer is always the same, "NVidea and ATI don't release any kind of source code for a driver for their cards." Of course, you don't have to support Windows drivers, you could support Linux drivers insted(NVidia at least has Linux drivers for most of their cards) but I still think that you would be best off supporting Windows drivers.

K.J.

I agree. What I'm saing is: As wingdows has a period of time when the hardrware companies write the drivers for it before it is released, this os will do the same. The driver system will work trough call gates. What u need is to load the code with the routines for a nvidia card for example and load tha adress in the call gate table. Thea would solve most of the problems as nvidia  will only use the hardware part to recompile and I will do the part of loading the driver and working it with the os. And yes, we'l first write the os and then request help t hardware companies. And yes I wrote drivers for Dos and MMURTL and didn't bother with windows as it sucks(no offence and I'm not trying to argue). MMurtl is an example os 32-bit multitasking, and I belive I know what it involves. And I don't just have a big view;
REMEMBER it all starts with a big view(a small view won't give big results, but a big view will give at least moderate results)

Try to give constructive criticism, not destructive. We don't wan't to make a windows clone; we want to make an os that has no flaws and that is by restructuring everything, not like microsoft did with their FAT16. They just patched it (made it slower) instead to restructure it and make a better one kind of like the NTFS(NT is a better os, too bad games don't quite work with it(direcx5?).

I once read, attached to the bottom of an internet guitar tab (Metallica),
"Design a system that even an fool can use, and only a fool will use it."

I don't have a problem with Windoze itself, only with being expected to use it.  The problem is with the windoze users, not windoze.  The only way you will sell an OS to the windoze using community is to make it simple enough to use it.  I use Linux mostly, but I also have a FreeBSD box.  I was completely Microshit free until a demand arose for a 'doze port of AGIL.

Trying to dummy down a real OS until your average windoze user can use it has been shown not to work.  Look at RedHat: Windoze users don't like it, it's too unix like, and Unix users don't like it, it's too 'doze like.

I hope you get what I'm trying to say.

I got your point. And I myself think that linux is too Unix like but I also think that windows is too **** like. I want an os that starts right up(doesn't boot for 2 years), like dos but runs a graphical user interface mixed(on the bottom) with a command line bar, so stupid people can use it but it can also be fast for smarter ones(fast commands). Kind of like AutoCAD.

I know what you mean by "an OS that starts right up".

I just let my Linux box run 24/7 because it takes sooooooo loooonnnngggg to boot(and shutdown also). I can get Win95 to boot in about 20 seconds with nothing in my startup group if I work on it long enough, WinXP boots in about 30 seconds to the logon screen, but then I haven't installed lots of software on it yet. I think that about 10-30 seconds for an OS boot is pretty good but I would prefer about 15 seconds or less. Instant on is basicaly impossible because that OS has to be loaded into memory .

K.J.

If you cut off all the unncesessary boot steps, and dont use as many stacks and GDTs and IDTs and LDTs that you'd eventualy erase and make new ones instead you could get an os that loads off a hard drive(7200rps) in 5-10 seconds. It's pretty amazing. The key word is simple. I think overcomplicated calling conventions and stuff like that might speed up the actual call but will slow down other stuff because it requires more code to be run so "simple" is the way to go. If the whole OS session would be faster because of a somewhat more complicated calling convention I'd rather give up 1-2 seconds at startup than have programs run slower.
Cheers,
Dori

My linux box boots like lightning with a small kernel.

I unfortunately need quite a large kernel (bzImage.)  The kernel decompression takes the longest by far.
This box is only a Celeron 333Mhz (hardly killer.)  I haven't actually timed it, but I reckon that the kernel (even the large one) boots in under 20 seconds (including decompression.) provided no hardware trouble is encountered.

Also starting services can take a long time (squid, bind, sendmail, etc.)  but this would be the same on any OS, as it it not an os function.

..

I'd have to disagree with that. The only reason I use microsoft windows now is that not only am I not a fan of macs but there are few gaming companies out there that cater to other OSs. For a lot of programs that originated off windows, apples and macs are still about half a year behind in the market. Sure windows is easy to use and has a very versatile interface, but that comes with the territory. Microsoft has the biggest monopoly on computer-related products there is. So it's natural that a lot of people use it. IBM's own OS/2 which I used for quite a while was very reliable, albeit slow on my upgraded 386 PS/2. Linux has a lot of resemblances to OS/2 so in part I can say I have gotten used to it pretty quickly. Although you can't get OS/2 on the public market anymore, Linux is a very good OS. It may not be very compatible, but that's because it has been overshadowed by Microsoft's own. The difference between the two OSs is that Linux is not easily corruptible. We have it's scarcity among consumers to thank for that.