osfree.org

#291 From: "sandervl2000" <sandervl2000@...>
Date: Sun Feb 24, 2002 4:25 pm
Subject: ReactOS / base for OS/2 compatible kernel (Was Re: NewOS) sandervl2000
Offline Offline
Send Email Send Email
Invite to Yahoo! 360° Invite to Yahoo! 360°


--- In osFree@y..., Jason Filby <jasonfilby@y...> wrote:
> >If your primary aim is to use NT/Win2k device drivers, then a far
> >easier solution would be to build a compatibility layer on top of
an
> >existing OS. That is possible as I have done this in OS/2 for Linux
> >drivers (SB Live) and for Win9x (a win9x audio driver I had to
> port).
> Most of our developers prefer the design of the Win NT kernel to
that
> of Linux.
So do I, but that's beside the point. If you want to run NT drivers,
build an emulation layor on top of your favourite OS.
If you want a great OS, why not design one yourself instead of copying
NT?
It's much more gratifying to design your own system than trying to
emulate an existing (and partly (largely?) undocumented) OS.
And the amount of work required will not be much larger. (including
a win32 emulation layer)

> >To run win32 applications you obviously don't have to build an NT
> >like operating system from scratch.
> All in all, the crazy people down at reactos.com think this is the
> best approach. Look at WINE -- even though they are getting really
> close to application compatibility, they'll never get 100% because
of
> trying to run win32 apps on Linux, which is a different OS design.
Nobody will ever be able to emulate the win32 api 100%. (even MS has
troubles with backwards compatibility) But you don't have to to get
good results.

Besides, you make it sound like NT is unique in its design. It's not.
Just an OS with a good design. Nothing more, nothing less.
NT has a nice kernel, but the user api is horrible. And MS keeps on
pushing more code into ring 0. Not exactly the best combination.

I'm not trying to bash ReactOS. If you guys can pull it off, then that
would be great. Just stating my opinion.

Sander

#292 From: "sandervl2000" <sandervl2000@...>
Date: Sun Feb 24, 2002 4:57 pm
Subject: ReactOS / base for OS/2 compatible kernel (Was Re: NewOS) sandervl2000
Offline Offline
Send Email Send Email
Invite to Yahoo! 360° Invite to Yahoo! 360°


--- In osFree@y..., "JMA" <mail@j...> wrote:
> >Returning to the osFree subject. I and some other programmers I
know
> >will never work on osFree or a derative of this project.A lot of
> >people on this list have strong opinions and ignore all legal
issues.
> Ignoring any legal aspect that I'm trying to get rid of *very* hard
!
>
> What stops you from participating ?
If you want to start a clean-room OS/2 project, then you have to
cut all ties with osFree. Start a new yahoo group, change the name
for starters.
You must avoid that at a later date people start to get suspicious
about where the code came from. Such rumours can really hurt a
project.
(even if completely false)

> >In my opinion the best option is to either use an existing mature
> >operating system (linux/freebsd) or write one from scratch that is
> >flexible enough to support multiple operating system personalities
> >(application & driver). The latter could for instance be based on
> >the l4ka micro kernel (http://l4ka.org)
> I agree that we should try to rip out what we want(*) and place it
> onto a suitable kernel. I would recomend exploring Linux since
> all its benefits.
Well, Linux is actually not my first choice. Personally I dislike unix
in all forms and Linux's threading might not be good enough for OS/2
emulation. (see a recent thread on c.o.o.programmer.misc)

But more important, why bother with an OS/2 subsystem for Linux? There
aren't that many great OS/2 applications. If you want OS/2 features
on Linux, then write a WPS-like shell (or enhance an existing one).
You can add an OS/2 API interface for Linux, but I don't expect many
people to use it for development.
IMHO it's basically a waste of time as you might as well migrate to
Linux completely.

> Supporting lots of operating systems using a personality layer
method
> is a great thing but its aiming to high *for just us*. Other groups
?
> are/should work with that, we should concentrate on the OS/2
> compatible personality.
Of course. All I'm saying is that you need to take into consideration
that mulitple personalities are a requirement when designing it.

Either you go the linux way and hope the OS/2 subsystem gets some
acceptance or you design a new OS (or take a non-unix kernel) and
rely on a win32 personality to run popular apps side by side with
your favourite OS/2 apps & the WPS.

Sander

#293 From: "ShadoW" <ShadoW.FmC@...>
Date: Sun Feb 24, 2002 5:01 pm
Subject: Re: NewOS shadow_fmc
Offline Offline
Send Email Send Email
Invite to Yahoo! 360° Invite to Yahoo! 360°


On Sun, 24 Feb 2002 13:52:40 +0100, JMA wrote:

>And your in a OS/2 group, just a fair warning

And we OS/2 users flame everybody who says NT without Win, because we are
zealots!
You wanted to say that, did you?

Sebastian

#294 From: Jason Filby <jasonfilby@...>
Date: Sun Feb 24, 2002 5:27 pm
Subject: Re: NewOS jasonfilby
Offline Offline
Send Email Send Email
Invite to Yahoo! 360° Invite to Yahoo! 360°


--- JMA <mail@...> wrote:
>On Sun, 24 Feb 2002 04:29:17 -0800 (PST), Jason Filby wrote:

>Thats not the point. There is a reason why Microsoft put this on
>their package.
>And your in a OS/2 group, just a fair warning

Yes. Its because they are very much legal aware... as are you, it
seems.

>Also, you are not building a Windows NT kernel - you are mimicing
>it, just like we would mimic the OS/2 kernel if we decided to go
that
>way. Unless you base your kernel on the real Windows NT kernel
source
>it will be as untested as anything else. You cannot do that, we
>cannot (and have little interest) of doing that.

I said the Win NT design is tested.

>>WINE is obvious for win32 -- reduce redundant coding. WinFree?
>>Never heard of it. Linux -- only for the internals of some
functions
>>_not_ for design.
>THIS TEXT IS FROM THE README FILE IN YOUR OWN SOURCE DISTRO.
>I DOWNLOADED IT TODAY !
> Something is wrong, I though you were the maintainer of ReactOS ??

Fair enough, you should have mentioned the source. The maintainer of
ReactOS? I am the project coordinator and an active coder. Please
look at how many files are in our source. I don't read through all of
them. This is not a project where one person controls every line of
code. Someone obviously someone saw fit to borrow from that project
and documented in the readme file. Big deal.

>While a do agree, get ready to be flamed
>PS: Dont even get close to the FreeOS yahoogroup ;--)

Flame away.

>Yes, but ReactOS still dont have that support in place, does it ?

No.

>Linux would be the best choice if I had anything to say. Its
>opensource, developed by hoards of people, there are mk distros
>and its quite mature.

Then do it already.

>Linux is mature, very well known, has everything we need though
>maybe not in a perfect way. It would be the perfect place to (at
>least test) implementing a OS/2 API layer.

Right. In the meantime someone is already looking to build an OS/2
subsystem on ReactOS. Perhaps more will join him in the future.

- Jason







__________________________________________________
Do You Yahoo!?
Yahoo! Sports - Coverage of the 2002 Olympic Games
http://sports.yahoo.com

#295 From: Jason Filby <jasonfilby@...>
Date: Sun Feb 24, 2002 5:33 pm
Subject: Re: ReactOS / base for OS/2 compatible kernel (Was Re: NewOS) jasonfilby
Offline Offline
Send Email Send Email
Invite to Yahoo! 360° Invite to Yahoo! 360°


--- sandervl2000 <sandervl2000@...> wrote:
>So do I, but that's beside the point. If you want to run NT
>drivers, build an emulation layor on top of your favourite OS.
>If you want a great OS, why not design one yourself instead of
>copying NT?

Drivers will work much better if the kernel is designed to support
them. The people at ReactOS think that Win NT has a good design.

>Besides, you make it sound like NT is unique in its design. It's
>not. Just an OS with a good design. Nothing more, nothing less.
>NT has a nice kernel, but the user api is horrible. And MS keeps on
>pushing more code into ring 0. Not exactly the best combination.

MS pushes code into ring 0 that, if it crashed, would be just as good
as if the kernel crashed. The GDI, for instance, is always used since
Win NT is a graphical OS. So why not have it in ring 0 for extra
performance? Besides -- the GDI, in fact, most of the code in ring 0
almost never crashes. Badly written drivers, for the most part, are
responsible.

>I'm not trying to bash ReactOS. If you guys can pull it off, then
>that would be great. Just stating my opinion.

Sure thing.

- Jason







__________________________________________________
Do You Yahoo!?
Yahoo! Sports - Coverage of the 2002 Olympic Games
http://sports.yahoo.com

#296 From: "JMA" <mail@...>
Date: Sun Feb 24, 2002 6:50 pm
Subject: Re: ReactOS / base for OS/2 compatible kernel (Was Re: NewOS) mailjmase
Offline Offline
Send Email Send Email
Invite to Yahoo! 360° Invite to Yahoo! 360°


On Sun, 24 Feb 2002 13:57:37 -0000, sandervl2000 wrote:

>--- In osFree@y..., "JMA" <mail@j...> wrote:
>> >Returning to the osFree subject. I and some other programmers I
>know
>> >will never work on osFree or a derative of this project.A lot of
>> >people on this list have strong opinions and ignore all legal
>issues.
>> Ignoring any legal aspect that I'm trying to get rid of *very* hard
>!
>>
>> What stops you from participating ?
>If you want to start a clean-room OS/2 project, then you have to
>cut all ties with osFree. Start a new yahoo group, change the name
>for starters.
>You must avoid that at a later date people start to get suspicious
>about where the code came from. Such rumours can really hurt a
>project.
>(even if completely false)
>
Sorry, but if I'm to believe people here this is no right.
No OS/2 clonage project can ever get away from that suspicion
if I'm to believe people like Eirik.

All I can do is :
a) set up absolute rules and
b) stand aside from the TPE distro

If I'm to be 100% sure I cannot hardly accept source from anyone,
can I ??

Can any project to be 100% clean accept code when the secret of
the "source leak" is out ??


>> >In my opinion the best option is to either use an existing mature
>> >operating system (linux/freebsd) or write one from scratch that is
>> >flexible enough to support multiple operating system personalities
>> >(application & driver). The latter could for instance be based on
>> >the l4ka micro kernel (http://l4ka.org)
>> I agree that we should try to rip out what we want(*) and place it
>> onto a suitable kernel. I would recomend exploring Linux since
>> all its benefits.
>Well, Linux is actually not my first choice. Personally I dislike unix
>in all forms and Linux's threading might not be good enough for OS/2
>emulation. (see a recent thread on c.o.o.programmer.misc)
>
>But more important, why bother with an OS/2 subsystem for Linux? There
>aren't that many great OS/2 applications. If you want OS/2 features
>on Linux, then write a WPS-like shell (or enhance an existing one).
>You can add an OS/2 API interface for Linux, but I don't expect many
>people to use it for development.
>IMHO it's basically a waste of time as you might as well migrate to
>Linux completely.
>
No, its a great place to try it out (an API layer) and see how hard
it would be. You can do that by writing specification but you and
I know that will never work.
It must be tried out.

And with Linux you can find out why it did not work.


>> Supporting lots of operating systems using a personality layer
>method
>> is a great thing but its aiming to high *for just us*. Other groups
>?
>> are/should work with that, we should concentrate on the OS/2
>> compatible personality.
>Of course. All I'm saying is that you need to take into consideration
>that mulitple personalities are a requirement when designing it.
>
Mulitple personalities may be a requirment for the kernel
the OS/2 clone layer is to run ontop but trying to do a WPOS
is nothing for the amount of developer the OS/2 community
could ever pull together. Atlest not as a opensource non
comersial project.

But starting another "build the best kernel" is idiotic. There are
so many such projects. Better find the right one and coop with
them.




Sincerely

JMA
Development and Consulting

John Martin , jma@...
==================================
Website: http://www.jma.se/
email: mail@...
Phone: 46-(0)70-6278410
==================================

#297 From: "JMA" <mail@...>
Date: Sun Feb 24, 2002 6:52 pm
Subject: Re: NewOS mailjmase
Offline Offline
Send Email Send Email
Invite to Yahoo! 360° Invite to Yahoo! 360°


On Sun, 24 Feb 2002 15:01:26 +0100 (CET), ShadoW wrote:

>On Sun, 24 Feb 2002 13:52:40 +0100, JMA wrote:
>
>>And your in a OS/2 group, just a fair warning
>
>And we OS/2 users flame everybody who says NT without Win, because we are
zealots!
>You wanted to say that, did you?
>
I am I'm I of the worst (VBG)

I do hope you know what and VGB means !






Sincerely

JMA
Development and Consulting

John Martin , jma@...
==================================
Website: http://www.jma.se/
email: mail@...
Phone: 46-(0)70-6278410
==================================

#298 From: "Lynn H. Maxson" <lmaxson@...>
Date: Sun Feb 24, 2002 6:53 pm
Subject: Re: OSFree and our future lynnmaxson
Offline Offline
Send Email Send Email
Invite to Yahoo! 360° Invite to Yahoo! 360°


Michal Necasek writes:
" Would you try to apply manufacturing processes to music, movies,
books?"

I'm making a second, hopefully shorter, response to this as I
infer that you think programming is some free form of
self-expression, thus the reference to such art forms in music,
movies, and books. In programming our basic goal lies in
providing a solution set which matches a given problem set, one in
which everything in the problem set is addressed in the solution
set and vice versa.

Moreover we seek to do it with minimal expression, i.e. writing.
No more, no less than necessary. When symbolic assembler was king
in the late 50's (yes, I was programming then) we expended great
effort in code reviews and competition as to whose solution
occupied the least space, minimal code. When we went to "higher"
level languages like Fortran, Cobol, or PL/I (which existed years
before C and its ilk) we gave up control over the underlying
assembly language produced, but nevertheless still competed for
minimal expression.

In doing this we often created spagetti code. Iverson in his
description of his specification language in his book "A
Programming Language" in 1962 which later became the APL
programming language treated a program as a vector in linear
space, exactly as it occurs in every computer. If you
"abstracted" out the pure sequential code, leaving only that which
may redirect control flow (decision and iteration), you could draw
the lines from either side of the vector, exit from a point and
its entry at another, and visibly see which drawn lines crossed
others: spagetti code.

If you do this using structured programming techniques, i.e. the
one-in/one-out control structures of sequence, iteration, and
decision, you will have no crossing lines, no spagetti code. You
will, however, see an increase in code size, number of lines of
code, as code is replicated (reused) in-line instead of
out-of-line as in a subroutine call.

Now there is no doubt that in-line code executes faster overall
than out-of-line due to the overhead of the invocation. So
out-of-line provides reuse, thus saving code space, but at a cost
in performance. In-line offers increased performance, but at a
cost in increased code space. Unless you use something like an
"include" statement reused code segments are not reusable, only
replicated.

The point is that you have the problem set which sets the
constraints on the solution set: the expression of one must match
the expression of the other, i.e. logical equivalence. You want
to have a solution set with minimal expression, i.e. writing, as
it is the writing time (and unfortunately the rewriting time)
which consumes the programming effort. All this is anathema to
any belief that programmers are engaged in some form of free-form,
self-expression.

You have two goals. One, an exact match between the solution set
and the problem set. Two, the ability to maintain that match as
the problem set changes. Now how you get that match may involve
some "creativity", i.e. new, not reused code. But if you are some
employer who has to pay for the coding effort or a vendor who has
to compete on the cost of the coding effort, you will make every
effort to maximise code reuse, thus minimizing additional coding
effort.

#299 Re: [osFree] NewOS
Expand Messages

JMA
Feb 24, 2002
On Sun, 24 Feb 2002 06:27:06 -0800 (PST), Jason Filby wrote:

>--- JMA <mail@...> wrote:
>>On Sun, 24 Feb 2002 04:29:17 -0800 (PST), Jason Filby wrote:
>
>>Thats not the point. There is a reason why Microsoft put this on
>>their package.
>>And your in a OS/2 group, just a fair warning
>
>Yes. Its because they are very much legal aware... as are you, it
>seems.
>

Hey people (not directed to you Jason) I hope you all know what a
"winkie" means.

And to be a little bit serious, yes, the amount of legalese here does
not help.

>>Linux is mature, very well known, has everything we need though
>>maybe not in a perfect way. It would be the perfect place to (at
>>least test) implementing a OS/2 API layer.
>
>Right. In the meantime someone is already looking to build an OS/2
>subsystem on ReactOS. Perhaps more will join him in the future.
>

Thats something I would like to endorse in this project.

I will not stop anyone from trying to build a osFree kernel. But I dont
see any reason, until proven otherwise, to make the API layer run
on many kernels.

I'd love our two projects to coop as long as your kernel design does
not limit our API/loader design as you dont get mad at us if we in
the future decides another kernel is the best for us.





Sincerely

JMA
Development and Consulting

John Martin , jma@...
==================================
Website: http://www.jma.se/
email: mail@...
Phone: 46-(0)70-6278410
==================================

#300 Re: [osFree] NewOS
Expand Messages

Jason Filby
Feb 24, 2002
--- JMA <mail@...> wrote:

>I will not stop anyone from trying to build a osFree kernel. But I
>dont see any reason, until proven otherwise, to make the API layer
>run on many kernels.
>I'd love our two projects to coop as long as your kernel design
>does not limit our API/loader design as you dont get mad at us if we
>in the future decides another kernel is the best for us.

Now we're getting somewhere
Ok when this group is set up to start coding, let us know. Then we
can share code whereever possible.

- Jason







__________________________________________________
Do You Yahoo!?
Yahoo! Sports - Coverage of the 2002 Olympic Games
http://sports.yahoo.com