On Tue, Oct 13, 1998 at 01:45:04PM -0400, Brian Ristuccia wrote:
>
> It wouldn't be bad at all, so long as it included source so that it could be
> patched to accomodate newer / different versions of the Linux kernel if the
> manufacturer went out of business or stopped supporting newer versions of
Well, this would be the only condition under which I'd put a
commercial driver with no substitute on my own system, but I don't see
how letting people make other choices is a bad thing.
> I'd hate to be stuck running kernel 2.0.35 two years from now because I
> bought some obsolecent device from a company that refuses to release a
> binary driver for a newer kernel version.
Then don't. However, in some cases, this would not be an issue.
There are some people still running a 1.2 kernel quite happily, mostly
because it suits their needs (and one of the folks I know is happy at
his record uptime, as well :).
> These commercial sound drivers are a real hassle, since the user must
[valid complaints and security issues elided]
> good hardware support is to Linux's success, I don't consider binary-only
> support good support at all. I'd hate to be stuck in Company X's position.
> I'm sure you'd feel the same way if it was your business on the line.
Us, sure. A lot of places, no. Look at the number of businesses
using MS. Obviously they aren't overly concerned about these things.
Heck, when security breaks under MS, for the most part, they don't
*get* a fix, broken drivers or not. But really, these are things best
left to the individual. If a person wants to take the risk, fine.
FWIW, I shelled out $20 for a commercial OSS license back when
OSS/Free didn't support PNP devices so I could get my AWE64 working,
and haven't regretted it. They've been remarkably punctual about
releasing drivers, even for the experimental kernels. So far, I still
haven't gone back to OSS/Free even though I've been told that the
problems long since resolved themselves, mostly because the drop-in
driver has been remarkably convenient, nifty features keep getting
added and I haven't had the time to play with ALSA. What did the use
of a non-free driver cost me this time? Nothing. I received the card
as a free replacement for an ancient SB classic (complete with the
mail-order plug-in replacement synthesizer chips, I forget the name)
when my brother needed me to do do some sound editing work for him
that required synchronous in/out. And if the company had vanished a
month later, leaving me out $20 for a module that probably would break
on the next kernel upgrade, well, really I wasn't out much, since the
card didn't work without booting to DOS first and then running loadlin
after the card initialized anyway. What did I gain? At a bare
minimum, several months of much reduced irritation. Hell, that's
worth my $20 right there.
The one valid argument that I've seen presented (and I'm surprised
you missed it, actually) was from RMS, and that is that when
commercial drivers become common, a lot of people lose the drive to
write free ones themselves and NDAs seem more reasonable, so allowing
commercial drivers to bind with the kernel could possibly leave the
free software community with fewer free drivers than would otherwise
have existed. I'm willing to grant that this is a possibility if
things progress, despite the fact that OSS doesn't seem to have
impeded the development of OSS/Free or ALSA, and the existence of
Xaccel doesn't seem to have slowed down the XFree group much. If you
want to fear commercial products, start there, rather than the fact
that commercial products frequently suck. Quite frankly, if companies
providing closed binary-only drivers don't provide decent support
for them, including a new version for every development kernel, I
don't think they're going to be much of a threat, since the general
populace is going to be too irritated to stop writing drivers (which
is really all that concerns us, I think), and other companies will
treat them about with as much support as they treat any other
unsupported product.
So frankly, I think we shouldn't worry about companies because
they might write poorly supported drivers. If you have to worry,
worry about companies that might write something really good, and
support it perfectly.
===========================================================================
Zed Pobre <zed@va.debian.org> | PGP key on servers, fingerprint on finger
===========================================================================
Attachment:
pgpYbwKGC7U0q.pgp
Description: PGP signature