Re: UDI

[Avery Pennarun <apenwarr@worldvisions.ca>]

On Wed, Sep 23, 1998 at 11:00:45AM -0700, Bill Moshier wrote:

> I see your point, but I disagree.  As new devices and drivers come out
> from Intel, SCO, and others, a conformance to UDI would potentially help
> both the ready acceptance of new hardware into the linux community, but
> also the acceptance of linux at the business level. Besides that, it would
> also help eliminate the 'fragmented Unix' concept that has given Microsoft
> the advantage for the last several years.

It would probably be relatively easy to support this UDI, once it exists, as
an optional way to support drivers.

The thing is, the Linux community has been very successful at making drivers
that work really well, as long as there's sufficient documentation
available.  For example, my Mach64 Rage3d video card does fewer strange
things in XFree86 than it does in Windows, and runs just as fast.  The Linux
ne2000 driver works with every ne2000 card, not only the one from my vendor. 
When I'm setting up systems, that makes my life a lot easier.

By supporting UDI, we allow companies to justify withholding their
programming information -- but don't worry, they'll supply a binary-only UDI
driver for their new card, and it'll probably work.  Too bad if there are
bugs, though.

Then there's the efficiency and cleanliness problem.  The Linux kernel
interfaces have been evolving for years -- even between 2.0.x releases, APIs
have been known to become incompatible.  Microsoft drivers change their
interface with every incarnation of Windows -- even Win95 and NT driver APIs
aren't the same.  Can we assume that Intel, SCO, and Sun, none of whose
products I particularly adore, will produce a UDI standard that's right the
first time?

Anyway, I'm not worried about it.  If UDI drivers start actually becoming
available, someone will make them work with the kernel.  It's not really
something Debian needs to do.  You might look at the
linux-kernel@vger.rutgers.edu mailing list.

Have fun,

Avery