Closed vs. open development methods (Was DVI output, ATI or nVidia)

Jonathan Dieter jdieter at gmail.com
Wed Jun 27 15:23:46 UTC 2007


On Wed, 2007-06-27 at 07:34 -0700, Lonni J Friedman wrote:
> On 6/27/07, Dr. Michael J. Chudobiak <mjc at avtechpulse.com> wrote:
<snip>
> > Is it good that the kernel bugzilla says "NO BINARY MODULES or other
> > tainted kernels. Do not file bugs here if you have any binary kernel
> > modules loaded, reproduce without that module first. NVIDIA users - THIS
> > MEANS YOU!"?
> >
> > In what way is the closed nvidia code and process better than being
> > open? How does that benefit users?
> 
> I never claimed that it was better.  I just said that its certainly no
> worse.  No one here has yet to provide any concrete evidence to prove
> otherwise.
> 
This is *unreal*!!  Are you actually claiming that the using a
closed-source development process is as good as an open-source
development process?  That there is no evidence to the contrary?

Let's look at two very different problems that I had when I got my new
laptop nine months ago.

The first problem:  The hard drive was very slow...so slow that Fedora
Core 5 would crash when I tried to install it.  I was also getting an
"Invalid MAP value" in dmesg.

The first solution:  I looked into the kernel source to find why I was
getting the Invalid MAP value.  It turned out that Intel's specs for my
hard drive controller say that the controller doesn't support 2 PATA + 2
SATA slots.  The ata_piix driver followed the specs.  I changed the
ata_piix driver (one or two lines of code) to allow said combination.
Everything started working perfectly.  Sent fix upstream (my first and
only kernel bugfix).  It was included in the next kernel update, and
made it (barely) into Fedora Core 6.  Woohoo!  Problem fixed after a
couple hours of work, plus an hour or so spent on sending e-mails
upstream.  And now nobody else will have this problem.

The second problem:  I couldn't get compiz to work with the NVIDIA card
that came with the laptop.  Once I enabled compiz, my system would work
fine for between one and ten minutes, and then the display would start
flickering and sooner or later the kernel would crash.

The second solution:  I report a bug using the nvnews forum, and get the
useful suggestion (from YOU) to "update my BIOS".  Except there isn't a
BIOS update available for my system.  And it's not a BIOS problem
anyway.  There's no way for me to track down the offending
code...because there's no source available.  All I can give you is some
"Xid" errors which obviously aren't useful to either of us.  Finally,
after days of searching, a friend from the forum points out a solution
that, according to you[1], "isn't possible in Linux for mobile GPU's
right now."  Except it does work...if I pass certain undocumented
parameters to the nvidia kernel module.

So, to sum up, with an open driver, I'm able to find the problem and fix
it (sending the fix upstream as well).  With a closed driver, I'm left
helplessly flailing about asking for help from what seems to be one or
two overworked NVIDIA employees and/or other people who are helplessly
flailing about.

Lonni, I don't want to be bashing you at all, because, from what I see,
you and one or two others are the only ones helping us Linux users on
nvnews.  But it really sucks that NVIDIA doesn't have a proper
bug-reporting tool for *all* their users, and that there's no way for us
to even *try* to fix our own problems.  And it's even worse that the
code isn't open so that those who do know what they're doing (myself
most definitely excluded) can improve it.

Our school is going to be buying 40 new computers this summer, and I've
specified Intel graphics because I'm tired of dealing with opaque driver
problems.

Jonathan

[1]http://www.nvnews.net/vbulletin/showthread.php?t=80888#13
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://listman.redhat.com/archives/fedora-list/attachments/20070627/94edbb2f/attachment-0001.sig>


More information about the fedora-list mailing list