December 04, 2005 Re: About writing a D OS - Titan Planning Committee | ||||
---|---|---|---|---|
| ||||
Posted in reply to clayasaurus | clayasaurus wrote:
>
> What is so bad about POSIX and *nix in general?
The POSIX spec is extremely vague just to ensure that almost any system could be made POSIX-compliant. The problem with tends to be that almost any application worth writing these days requires specific knowledge or behavior that POSIX defines as optional or simply doesn't guarantee. Multithreading is a perfect example here, as multithreading is essentially an optional feature in POSIX and even when supported the POSIX spec requires mutually exclusive access to *all* shared data for the program to avoid undefined behavior. I think the features that POSIX outlines are pretty good for the most part, and the spec includes some incisive observations, but I wouldn't want it as the basis for a completely new OS design. That said, POSIX support allows for compatibility with a tremendous amount of pre-existing code, so supporting POSIX can be quite a good thing, just so long as you don't stop there. BeOS actually had a great deal of promise as a new OS design with POSIX support before Be, Inc. gave up on the desktop OS in favor of one targeted at internet-connected kiosks. That was the kiss of death for them, though their bottom line was pretty grim either way.
Sean
|
December 04, 2005 Re: About writing a D OS - Titan Planning Committee | ||||
---|---|---|---|---|
| ||||
Posted in reply to John Reimer | John Reimer wrote: > > One thing I liked about it over Windows-based PC's was the use of proper names for devices. I can't believe we're still using the nasty A:, B:, and C: nomenclature for drive devices within the Windows world (as inherited from its ancient DOS roots). Definately! That we're *still* stuck with drive letters is completely absurd considering that Unix has had volume mounting for something like 30 years now. The Windows filesystem is one of the things I like least about the OS these days. Though apparently the folks at MS agree, since they're now working on a filesystem a lot like BeFS for whatever follows Vista. > Unix, for all it's copious administration complexity, certainly had consistency to it's credit: its "everything-is-a-file" mentality makes for a "clean" and logical system. > > If similar, yet original, ideas could be carried over to a new D-based OS, that would be great. Agreed. Sean |
December 04, 2005 Re: About writing a D OS | ||||
---|---|---|---|---|
| ||||
Posted in reply to Chris | "Chris" <ctlajoie@yahoo.com> wrote in message news:a8deea61387a8c7c6fec0a52dcc@news.digitalmars.com... > I really don't have very many ideas regarding the OS. I am just trying to be realistic about all of this recent ambition of creating an OS in D. one of the big problems with making a new OS is compatibility. If you want your > OS to be suitable for typical PC users (I gather this is the intent of "Apollo") > then it would need to be compatible with existing software (including drivers, > with little modification). Realistically, you can't create new software that > does everything (or even close) you can do in Windows or Linux. > Also, you have to generate a huge amount of interest in your OS to get people > to donate their time to your cause. > I *really* don't want to discourage innovation. I just want to make sure > it isn't misplaced. But if you have enough passion for creating this, it > can definately work. SkyOS comes to mind. One way that Win32 took off was that it would run, unmodified, MS-DOS executables. Any new OS will have a much better chance of success if there's at least some level of binary compatibility. For example, DMD itself uses only a handful of basic, simple OS API's. If the new OS were to support the Windows PE executable format, and supply those basic API's, quite a lot of Win32 command line tools will run under the new OS without needing recompilation. This would be a big help in bootstrapping the OS. |
December 04, 2005 Re: About writing a D OS | ||||
---|---|---|---|---|
| ||||
Posted in reply to Walter Bright | Walter Bright wrote:
> One way that Win32 took off was that it would run, unmodified, MS-DOS
> executables. Any new OS will have a much better chance of success if there's
> at least some level of binary compatibility. For example, DMD itself uses
> only a handful of basic, simple OS API's. If the new OS were to support the
> Windows PE executable format, and supply those basic API's, quite a lot of
> Win32 command line tools will run under the new OS without needing
> recompilation.
>
> This would be a big help in bootstrapping the OS.
>
>
Very interesting. That's a good consideration!
-JJR
|
December 04, 2005 Re: About writing a D OS | ||||
---|---|---|---|---|
| ||||
Posted in reply to Chris | In article <a8deea61387a8c7c6fec0a52dcc@news.digitalmars.com>, Chris says... > >I really don't have very many ideas regarding the OS. I am just trying to be realistic about all of this recent ambition of creating an OS in D. one of the big problems with making a new OS is compatibility. If you want your OS to be suitable for typical PC users (I gather this is the intent of "Apollo") then it would need to be compatible with existing software (including drivers, with little modification). Realistically, you can't create new software that does everything (or even close) you can do in Windows or Linux. tot You are totally right in that point. I know Trevor first had the idea to make an exokernel, which, if I remember well, was working with module. Example, there was a windows module which you need to have program compatibility, etc for other ones... howevor, I'm not sure this way of proceeding work for drivers; Trevor would be better to answer this point. (BTW, sorry Trevor if I did a mistake) Pierre-Luc Cyr |
December 05, 2005 Re: About writing a D OS - Titan Planning Committee | ||||
---|---|---|---|---|
| ||||
Posted in reply to Sean Kelly | In article <dmvulu$259f$2@digitaldaemon.com>, Sean Kelly says... > >John Reimer wrote: >> >> One thing I liked about it over Windows-based PC's was the use of proper names for devices. I can't believe we're still using the nasty A:, B:, and C: nomenclature for drive devices within the Windows world (as inherited from its ancient DOS roots). > >Definately! That we're *still* stuck with drive letters is completely absurd considering that Unix has had volume mounting for something like 30 years now. The Windows filesystem is one of the things I like least about the OS these days. Though apparently the folks at MS agree, since they're now working on a filesystem a lot like BeFS for whatever follows Vista. While you're on the subject, there's something else to be learned from Vista: Monad. The idea that you can elevate the shell to an interactive script environment isn't 100% new, but the idea that all applications are (scriptable) *objects* is. Suddenly, the shell stops becomming this exception to where the "real" work is done (inside of applciations), so the entire system becomes much more homogenous... just like unix back in the day. To expound on the idea: where text files and streams were the "lingua franca" of *nix, the way forward could be accomplished with some sort of universal tabular format (binary or XML) and object id's. - EricAnderton at yahoo |
December 05, 2005 Re: About writing a D OS - Titan Planning Committee | ||||
---|---|---|---|---|
| ||||
Posted in reply to Hytak | > Haiku-os is a project by BeOS fans to make an open-source revival of BeOS... http://haiku-os.org/learn.php
How about DeOS :)
|
December 05, 2005 Re: About writing a D OS - Titan Planning Committee | ||||
---|---|---|---|---|
| ||||
Posted in reply to Hytak | Hytak wrote:
>> And I'd say...
>>
>> Please don't fall into the win32 swamp! Please, please don't use WinNT as your foundation!!!!
>>
>> -JJR
>
>> In article <dmvi84$1oji$1@digitaldaemon.com>, Hytak says...
>>
>> I agree too, moreover, I heard good comments about BeOS although I never tried
>> it.
>
>> Never used it either, but I too heard they had a pretty good architecture. And
>> since they were targeting multimedia stuff, I assume it was pretty fast too.
It was. And built from the ground up to be pervasively multithreaded. I remember seeing an example where someone had ~20 instances of a Star Wars trailer running on their desktop and none of them dropped a frame... even when the user dragged windows around the screen. You could also do stuff like move mp3 files around the filesystem while they were being played and not only would the OS let you do it, but it had no effect on playback quality or OS response (try that with Windows). The only weird thing about BeOS was that everything worked via message passing... and there were no hard guarantees that a messages would be delivered, even if they were just being sent between controls in an application. This made UI programming a bit awkward--I'm not sure if it was addressed in a later version of BeOS or not. BeOS also lacked a multi-user interface and the network layer was garbage. BeOS did get a huge tuneup after they refocused on kiosks however, but I don't think that version ever made it to the PC market. I used BeOS for a few years and finally gave up on it when they stopped releasing PC updates.
Sean
|
December 05, 2005 Re: About writing a D OS - Titan Planning Committee | ||||
---|---|---|---|---|
| ||||
Posted in reply to Sean Kelly | Sean Kelly wrote:
> Hytak wrote:
>>> And I'd say...
>>>
>>> Please don't fall into the win32 swamp! Please, please don't use WinNT as your foundation!!!!
>>>
>>> -JJR
>>
>>> In article <dmvi84$1oji$1@digitaldaemon.com>, Hytak says...
>>>
>>> I agree too, moreover, I heard good comments about BeOS although I never tried
>>> it.
>>
>>> Never used it either, but I too heard they had a pretty good architecture. And
>>> since they were targeting multimedia stuff, I assume it was pretty fast too.
>
> It was. And built from the ground up to be pervasively multithreaded. I remember seeing an example where someone had ~20 instances of a Star Wars trailer running on their desktop and none of them dropped a frame... even when the user dragged windows around the screen. You could also do stuff like move mp3 files around the filesystem while they were being played and not only would the OS let you do it, but it had no effect on playback quality or OS response (try that with Windows). The only weird thing about BeOS was that everything worked via message passing... and there were no hard guarantees that a messages would be delivered, even if they were just being sent between controls in an application. This made UI programming a bit awkward--I'm not sure if it was addressed in a later version of BeOS or not. BeOS also lacked a multi-user interface and the network layer was garbage. BeOS did get a huge tuneup after they refocused on kiosks however, but I don't think that version ever made it to the PC market. I used BeOS for a few years and finally gave up on it when they stopped releasing PC updates.
>
>
> Sean
I partook of the minor BeOS craze when it first came out. I bought the very first x86 release despite strong warnings by the company that it ran on very select hardware (Matrox video card, certain motherboards etc).
Apparently the tuneup you refer to managed to surface publicly in the form of ZETA OS by YellowTAB (www.yellowtab.com) They somehow managed to license parts of the unreleased BeOS 6 (from PalmSource? Nobody seems to know how they did that... there have been accusations...) and build on the BeOS tradition. The unreleased Beta 6 supposedly had a completely recoded network stack that was very fast. At the time, it was said that there was hardware accelerated OpenGL support for Radeon Graphics cards. I remember really looking forward to version 6 coming out. Unfortunately, it never did.
I don't know how well ZETA sells, but it's interesting, nonetheless.
-JJR
|
December 05, 2005 Re: About writing a D OS - Titan Planning Committee | ||||
---|---|---|---|---|
| ||||
Posted in reply to John Reimer | In article <dn12k1$ejb$1@digitaldaemon.com>, John Reimer says... > > I remember really looking forward to version 6 coming out. >Unfortunately, it never did. > >-JJR YES - BeOS was awesome - I still have a Max edition I install on machines all the time for fun. It's sooo efficient that an otherwise usless doorstop of a pentium box can become a fun toy again. The saddest part about BeOS being sold to palm, is .. What the hell did palm do with it / need it for? I mean, hey - buy the company, tweak the code, go somewhere with it... I was pretty pissed when 6 never came out too. Anyhoo - I am not against writing a POSIX compatible OS Library for the Titan Kernel, (an exokernel) but I don't plan to use ANY POSIX standards in writing the kernel itself.. I think POSIX has it's time and place, and this kernel is not that place, and now is not that time. Glad to see many of you loved BeOS too.. Keep those ideas coming - the website seems to be working out well. Feel free to post there too. http://www.sequenceartworks.com/labs/titan/index.php Thanks, Trevor Parscal |
Copyright © 1999-2021 by the D Language Foundation