Absolutely, this is exactly what I had in mind. The project needs a solidly working solution right now, even at some expense of speed (though it's arguable). It is very hard to properly rewrite our existing win32 subsystem without introducing regressions, and by the time this is done, the world may forget Win32 exists.

Arwinss gives key advantages to this:
1. It is possible to start a win32 subsystem with the proper, better design without dooming the project for 10 more years (existing win32k started in 1999) of work without any realworld useful result.
2. It is possible too make a new win32 subsystem aiming at future, not at present (e.g. use C++ if that fits better, use more modern architecture if needed, etc).
3. It is possible to gain new developers to do this task because with Arwinss ReactOS is going to get real usage, and come to a new height. Just imagine all good from Wine (compatibility) without all bad parts (Linux/XWindows related).
4. It allows to fix problems in other parts of the system (arwinss already uncovered bugs in csrss for example, more to come). Developing a new win32 susbsystem against bad users will result only in more pain (James knows how fun was it to see desktop class becoming local).


WBR,
Aleksey Bragin.


On Jul 29, 2009, at 3:38 PM, Ged wrote:

What people should also know is that if this branch ever does make it into trunk it will only be used as a temporary solution until the correct implementation is ready. This is by no means a permanent solution!

What it will do is act as a temp replacement which will hold things together and allow work on the real subsystem to accelerate.

 

At the moment the current subsystem must be kept stable as it’s our main component and needs to stay regression free whilst rewriting major parts to make it more compatible. Not an easy task!

If the Arwinss model can take over for a while it will gives the win32 subsystem developers breathing space to rewrite / hack / break / fight / kill / molest and eventually improve the real implementation without worrying about breaking reactos for everyone else.

 

In the long run this may be a great solution to improve the compatibility and stability of the real reactos win32 subsystem.

 

Ged.

 

From: Timo Kreuzer [mailto:timo.kreuzer@web.de]
Sent: 29 July 2009 11:33
To: ReactOS Development List
Subject: Re: [ros-dev] Arwinss architecture

 

Oh, c'mon, people.

This diagram says nothing.
Has anyone of you even a clue of how it would look like for Windows?
The only difference to Windows design that can be seen from this diagram is the addition of the NT driver and X11 driver.
What it doesn't show is where which parts of the subsystem are located. And that would probably show compatibility problems and bad performace.

Regards,
Timo

Brian schrieb: 

i cant help but feel that is how it should have been done in the first place
by the way the diagram looks
 
 
 
  
 



 
_______________________________________________
Ros-dev mailing list
Ros-dev@reactos.org
http://www.reactos.org/mailman/listinfo/ros-dev

 

_______________________________________________
Ros-dev mailing list