On Wed, Apr 26, 2000 at 11:20:08PM +0200, Alexander Langer wrote:
>BTW: I'm also satisfied with binary-only drivers, but as read in this
>thread, some guys would not be satisfied with this, too, what I myself
>can't understand.

> Thus spake Thomas Michael Wanka (tm_wanka@earthling.net):
> > What do you do if the manufacturer no longer supports your piece of 
> > hardware? If the driver will not work with the next version of your OS you may 
> > keep with the old version of the OS or by new hardware. Having the sources 
> > gives you the option to adapt the driver or have it done by someone else.
> 
> Don't you understand that?
> If there is _no_ driver, you never will be confrontated with the
> problem, that the manufacturer stops supporting your OS.
> If there is a binary one, you can be happy for the time it is
> supported.
> _Of course_, I prefer OpenSource drivers, but before I don't use this
> piece of hardware I stick to binary ones.

1) Ongoing support. I've replaced scsi and audio cards for people, in their
	windows machines -> when the old card manufacturer went bankrupt, 
	they don't produce new drivers anymore. And just because it's an
	old/slow card, doesn't mean it's useless, when a modern card is
	expensive, and you have something you want to experiment with.

2) Security. You like running kernel code that hasn't been subjected to
	public audit? Just wait for the announcement someday... Linux
	tape backup device driver LKM provides back-door root login...

3) Security again. Unless I'm testing something LKM related, my kernels
	have no LKM support compiled in. Even if you trust your commercial
	driver authors (who are too embarassed to let you see their code?)
	not to trojan your system - LKMs allow someone who does compromise
	your system to become completely invisible, and replace/intercept
	parts of your kernel! System compromises should be considered
	inevitable. Someday, a hacker is going to read about the latest hole
	before you do. Yes, you can reduce the odds, stay on top of Bugtraq
	etc, but you can't make it 0%. If you knew no one could ever break
	in, you wouldn't need to run things like 'tripwire' either. 

( See: "Attacking FreeBSD with Kernel Modules" for an example.
  http://thc.inferno.tusculum.edu/files/thc/bsdkern.html )

4) Platform support. So, what are the odds that PCI card vendor 'X' is
	going to supply a binary (LKM) driver for all *twelve* platforms
	that NetBSD runs on which might have a PCI bus (By my count,
	Alpha/Amiga/arm32/Atari/BeBox/Cobalt/hpcmips/i386/macppc/ofppc/sh3/
	sparc64 - though I'm not certain of PCI on arm32/sh3/hpcmips)
	One properly written driver available as source in NetBSD will
	work on all of the above. Don't fight to turn companies on to
	Lin-tel for profit reasons, because then you'll be stuck with
	Intel Linux, just like Intel Windows. Encourage companies to
	decide whether they're in the business of selling hardware, or
	writing drivers.

-- 
David Maxwell, david@vex.net|david@maxwell.net --> Mastery of UNIX, like
mastery of language, offers real freedom. The price of freedom is always dear,
but there's no substitute. Personally, I'd rather pay for my freedom than live
in a bitmapped, pop-up-happy dungeon like NT. - Thomas Scoville