Call me a total geek, but when I first read this thread, I got really excited. The basic premise of the mentioned project , called the "Linux Unified Kernel" is to use Wine to natively support many standard windows system calls, effectively taking much of Wine out of userspace and into kernel land. Dan Kegel's brief explanation:
It looks like they added hooks to the linux kernel to accept windows nt syscalls. Maybe they even allow using the system's normal shared library loader instead of Wine's special one. This is something I've often wanted to do, but it was way lower priority than getting wine working. I haven't looked at their project at all, no idea if it was done well.
There are some obvious benefits to this (some speed improvements being some) and some obvious pitfalls (this becomes very linux-only whereas now Wine currently supports Linux, BSD and Mac). I was a bit concerned that such a project would never take off due to lack of upstream (Kernel) acceptance. However, Dan Kegel apparently has some insider info which floored me:
We would, I think, like to move wineserver into the kernel sometime. It's been discussed before. Linus is not opposed to having native support for win32 system calls. A fellow at Redhat wrote a kernel module for wine a number of years ago, but it just wasn't time yet.
[^] # Re: mais
Posté par abramov_MS . En réponse à la dépêche La guerre des formats de bureautique normalisés ISO commence. Évalué à 2.
http://www.winehq.org/?issue=342#A%20Wine%20vs.%20Vista%20co(...)
Call me a total geek, but when I first read this thread, I got really excited. The basic premise of the mentioned project , called the "Linux Unified Kernel" is to use Wine to natively support many standard windows system calls, effectively taking much of Wine out of userspace and into kernel land. Dan Kegel's brief explanation:
It looks like they added hooks to the linux kernel to accept windows nt syscalls. Maybe they even allow using the system's normal shared library loader instead of Wine's special one. This is something I've often wanted to do, but it was way lower priority than getting wine working. I haven't looked at their project at all, no idea if it was done well.
There are some obvious benefits to this (some speed improvements being some) and some obvious pitfalls (this becomes very linux-only whereas now Wine currently supports Linux, BSD and Mac). I was a bit concerned that such a project would never take off due to lack of upstream (Kernel) acceptance. However, Dan Kegel apparently has some insider info which floored me:
We would, I think, like to move wineserver into the kernel sometime. It's been discussed before. Linus is not opposed to having native support for win32 system calls. A fellow at Redhat wrote a kernel module for wine a number of years ago, but it just wasn't time yet.
For a related BSD thingy, see http://www.kernel-traffic.org/wine/wn20010313_87.html#1 That project was still going as of 2006, I think, but didn't get terribly far.