From: Jon M. Taylor <taylorj@gaia.ecs.csus.edu>
  To  : ggi-develop@eskimo.com
  Date: Fri, 9 Apr 1999 16:30:15 -0700 (PDT)

Re: UDI proposed specs released for review

On Thu, 8 Apr 1999, Joseph Carter wrote:

> On Thu, Apr 08, 1999 at 02:37:00PM -0700, Jon M. Taylor wrote:
> > > In case any of you missed the announce on slashdot:
> > > 
> > > http://www.sco.com/UDI/f-specs.html
> > > 
> > > Grab them while the're there.
> > 
> > 	I did just that.  The core spec file is over 400 pages, whew... I 
> > like the 'metalanguage' concept, it is somewhat like the meta-register 
> > idea I was tossing around a few months back, but about 1000 times as 
> > complex.  I hope it all works out well, because I'll probably have to 
> > start using it soon if it gets adopted widely which looks likely.
> 
> Adopted by whom?  

	Industry.  The people who will soon be writing most Linux drivers.

> I don't think we'll see this thing in Linux real soon. 

	We don't have to get it into Linux proper - a module that implements
the core spec and which the driver modules are loaded on top of should be
sufficient. 

> Linus didn't seem to like the idea at all the last time this came up and
> despite the opinions of some on this list I will agree with Alan and say
> that binary only drivers in the kernel are a BAD thing.  

	They won't have to be 'in the kernel'.

> Yet the UDI spec
> is all about making drivers source compatible all around and binary
> compatible on like architectures.
> 
> If M$ adopts this thing, all new drivers written for windoze will run on
> Linux.  The hardware vendors would have to do nothing in order to stamp
> "Linux Compatible" on their hardware.

	WHy is that a bad thing?

> If that doesn't happen (it probably won't) we still have to convince
> vendors to write drivers for us or write them ourselves.  Only difference
> is that the vendors will have a handy way to write binary drivers for
> Linux without having to worry about keeping them up to date with the
> kernel (the primary thing keeping them from doing this now is that they
> don't want to try and keep up with kernel development!)

	Correct.  In fact, mini-UDI type things will have to be written 
anyway if UDI falls through for some reason.

> I don't think we should allow proprietary unices to shape Linux into
> another proprietary unix.  

	As I have pointed out, there is no 'allowing' of anything here. 
Binary-only modules of various kinds will be released, and some of those will
likely be UDI-related.  That is what _will_ happen.  Linus would have to
remove modules from the kernel entirely to avoid this, and he won't do that
because this type of situation is exactly why he added modules in the first
place - so people could do their own thing with the kernel without having to
bother him or affect the core kernel design. 

> If they succeed they may not win, but we are
> sure likely to lose.

	I disagree.
 
> Debian is very strongly opposed to binary only drivers, and so am I.

	I would _prefer_ to see Creative and other hardware manufacturers
open their specs.  But as long as they are willing to produce their own
drivers, the fact that they are binary-only really doesn't bother me all that
much.  And it didn't bother me even before I went to work for Creative, if
you are wondering about that.

Jon

---
'Cloning and the reprogramming of DNA is the first serious step in 
becoming one with God.'
	- Scientist G. Richard Seed