>>> Its name is Project UDI. The main objective of this effort is
>>> to provide an interface to the device drivers that is common
>>> to all UDI compliant OS's. That means that to port a driver
>>> from one OS to the other, you don't need to _change_ the
>>> source code, but just compile it in the new OS's compiler. I
>>> believe that this is a great initiative and I think that Linux
>>> could be inside that group.
For those that don't know about it yet:
http://www.sco.com/products/layered/develop/devspecs/udi/index.html
>> Most drivers are largely OS specific code (the actual hardware
>> fondling is in the minority). Abstracting that in a meaningful way to
>> a meaningful range of operating systems is unlikely to succeed, unless
>> you don't mind having no performance at all.
Low-performance vs. no-performance, hmmm?
> I also wrote/maintained drivers for several OS's (including SCO
> Openserver, Unixware, BSDI, FreeBSD and, of course, Linux) and there is
> one thing that is what makes the port withou source code changes
> impossible: the incompatibility between the OS's driver-kernel interfaces.
The kernel can provide whatever API is needed. Wrappers can be used.
> Even though the idea of creating a common interface to all OS's seems
> feasible, you have mentioned a very important issue to consider:
> performance. I don't even dare to talk about that anymore. Great point.
Not every driver is for FDDI on a 486. At the very least, the UDI
drivers could support a development system long enough to write
a normal Linux driver.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu