[ardour-users] ardour-users Digest, Vol 37, Issue 16

[Thomas Vecchione]

> What I was suggesting is that the debate as to whether the distribution
> uses source code,  tar balls,  rpm,  or deb packages or finds a driver
> acceptable is moot without the initial delivery.  

The debate was why manufacturers stay away from Linux.  I was pointing 
out that what a manufacturer does to provide a driver has nothing to do 
with the problems observed by the poster, not whether or not enough 
drivers are provided by the manufacturer.  I fully agree the 
manufacturer should be providing the driver, and that many are not.

> Your right that the distributions probably would have issues in
> packaging them due to quality issues,  copyrights,  and other reasons. 
> 

Copyrights should never be an issue for an open source driver submitted 
to the kernel tree for use with hardware.  The majority of drivers in 
Linux fit this fine.  Even those questionable drivers that exist(NVidia) 
have at least an open source connection layer that is still licensed 
under the GPLv2 I believe.

> 
> To often somehow somewhere the ball has been dropped as to distribution.
> Its bloody evident when you read the install instructions which wander
> all around and sometimes have language barriers.  Bless those who tried,
> but its not the same. So what are these drivers then?  Ghost ware?  
> 

As I mentioned earlier, for anyone there is a standard installation 
procedure that will work on ANY distribution.  But in doing so you may 
break the package management on that distribution.  It may work fine, 
but it depends on whether that package management ever tries to 
overwrite the driver.  In either case though it would hardly be ghost ware.

> I just got to wondering recently how many drivers are indeed out there
> without the help of the companies and how many of them are just in need
> of a little validation before they work fine and are fairly easy to
> install.  The best would be a proper package for each distribution.
> 
> There must be a significant amount of them,  and not just audio cards.
> Audio cards are fairly easy to track,  but something more obtuse
> probably is sitting on someones personal web site out there.  

The vast majority of them are in the kernel tree.  They have to be 
compiled against the GPLv2 licensed kernel headers, which requires that 
they be licensed similarly, and as a result they tend to be merged into 
the kernel tree.

How often do you have to go to a manufacturer's website for a driver in 
linux?  The only cases I can think of are in the case of ATi and NVidia 
should you choose to use those closed source drivers.  However even in 
their cases many distributions provide packages that assist in the 
install so you don't have to go to the website to get it.  Also I 
believe there are now two promising projects, one for each manufacturer, 
for getting the open source equivalents up to speed in supporting 3D 
acceleration on the newer cards.  I can't remember the ATi one but I 
believe the NVidia is called Nouveau.

> The next question would be why would a distribution want to carry all of
> the drivers?   If there was a repository with both open source (linux
> user made)  and proprietary drivers that did testing, provided the
> standard, the problem might be be solved,  if people let it.  

Because the distribution typically compiles the kernel, and in the case 
of the vast majority of drivers out there(I can't think of any userspace 
off hand, but I suppose there might be) you need to have the headers of 
the kernel that was compiled, in order to compile the drivers against it.

Thus there really is no way around that unless every distribution 
compiles the drivers on install, which many people stay away from Gentoo 
for the reason of they don't want to compile everything, even if it is 
automated.

> Packagers would just package them up and send them back for inclusion
> and distributions would just link to it to help make their downloads
> less.  Of course the debate your talking about,  then has merit.  
> 

That is what happens now.  The packagers for that distribution make the 
package of the driver, which consists of the binary compilation of the 
source code, compiled against the kernel that that distribution uses.

The source code for at least 90 percent of the drivers used in linux, if 
not more, is maintained as part of the kernel tree.  When you build a 
distribution you are getting the code from the kernel tree.

> It seems to me that hardware manufacturers are a bit like spoiled brats
> and there isn't a Linux equivalent of Windows Hardware Quality Labs to
> submit their drivers to.

That is what the kernel maintainers do.  You submit your driver to the 
kernel tree, the maintainers of the kernel look it over give feedback, 
test it, and if it meets standards merge it into the tree.  There are 
several branches of the kernel maintained by different devs specifically 
for this purpose.  See how long it took to get Ingo's patches into the 
kernel as an example.  Or the Xen drivers?  KVM got in fairly quickly on 
the other hand.

> The other interesting thing that develops is companies who are nicely in
> bed with MS and Apple can quietly release code to one channel and not
> several scattered developers who are hopeful to develop something.  The
> middle channel is a way of life for them,  naturally they relate. 
> 

See above.  In releasing code as open source, it naturally means that it 
HAS to be availiable to anyone.  The kernel is maintained by a 
collection of developers, not just a handful scattered across the globe, 
but a many employed by a variety of companies, along with many that 
donate their time.  You release an open source driver, you are not 
depending on others to develop it, you are developing it yourself in the 
same amount of time as you would for any other OS, and maintaining it in 
the same amount of time as you would for any other OS.

That aside the kernel devs I believe sent out an open letter recently 
offering to develop the driver for free if information on the hardware 
suitable to develop it from is submitted, and the OSDL I believe it was 
even offered to handle nessecary NDAs.  I don't know the details of it 
maybe someone else around is a bit more familiar with it, they can feel 
free to jump in and provide the details as needed.

> Sorry to seem to give you a hard time on it.  Having feel into a gap
> between a distribution and firmware lately,   I suppose this has been on
> my mind.  No offence meant.   

None taken, but I can't help but think you are still misunderstanding a 
few things, or I am completely missing what you are trying to say.


         Seablade