What are your thoughts, conserns, and recomendations on the development of an exo-kernel, or a chache kernel.
Exo-kernel research:
http://amsterdam.lcs.mit.edu/exo/
cache-kernel research:
http://www-dsg.stanford.edu/papers/cach ... /main.html
Thoughts on an exo/chache-kernel
-
Pype.Clicker
- Member
- Posts: 5964
- Joined: Wed Oct 18, 2006 2:31 am
- Location: In a galaxy, far, far away
- Contact:
Re:Thoughts on an exo/chache-kernel
from what i've been reading so far about them, exo-kernels do not really satisfy me. Maybe i need to read more about it, but roughly, saying that the application is going to handle protocol stacks and file systems sounds a bit like MS-DOS days.
I'm not saying that it's wrong to do so, but it seems that it will lead to limited possibilities of applications interoperations. I mean, sharing of hardware resource will quickly become complex, and it'll be a nuisance if every application have to deal with it.
The approach of having a more complex OS as a library is a possible solution, but how can instance of the OS-lib in application X synchronize its decisions with decision of the same OS-lib in application Y ?
Imho, pure exo-kernels sounds to be restricted to single-targetted computers (Internet Servers, appliances, gaming consoles, etc), but i would be very surprised if it could handle a complete multimedia system. It would be like trying to have just DirectX as the user interface to the system, and nothing like window manager or common GUI.
I'm not telling it worth nothing, just that it doesn't sound appliable as-is to desktop multimedia computing.
Now, i'm less comfortable with cache kernels (actually, i haven't heard about them before today), but it doesn't sounds more efficient than microkernels as calls to the "application kernel" still need to go through the "cache kernel" (system calls, page faults, etc)...
I'm not saying that it's wrong to do so, but it seems that it will lead to limited possibilities of applications interoperations. I mean, sharing of hardware resource will quickly become complex, and it'll be a nuisance if every application have to deal with it.
The approach of having a more complex OS as a library is a possible solution, but how can instance of the OS-lib in application X synchronize its decisions with decision of the same OS-lib in application Y ?
Imho, pure exo-kernels sounds to be restricted to single-targetted computers (Internet Servers, appliances, gaming consoles, etc), but i would be very surprised if it could handle a complete multimedia system. It would be like trying to have just DirectX as the user interface to the system, and nothing like window manager or common GUI.
I'm not telling it worth nothing, just that it doesn't sound appliable as-is to desktop multimedia computing.
Now, i'm less comfortable with cache kernels (actually, i haven't heard about them before today), but it doesn't sounds more efficient than microkernels as calls to the "application kernel" still need to go through the "cache kernel" (system calls, page faults, etc)...