Re: [PATCH 0/3] New firewire stack

["Kristian Høgsberg" <krh@redhat.com>]

Stefan Richter wrote:
...
>> Another point is the various streaming drivers.  There used to be 5 different
>> userspace streaming APIs in the linux1394: raw1394, video1394, amdtp, dv1394
>> and rawiso.  Recently, amdtp (audio streaming) has been removed, since with
>> the rawiso interface, this can be done in userspace.  However the remaining 4
>> interfaces have slightly disjoint feature sets and can't really be phased out.
> 
> The old iso API of (lib)raw1394 has been marked deprecated and
> undocumented in libraw1394's documentation for some time, and will go
> away in 2007.
> 
> Dv1394 might go away in 2007 too if there is enough effort to move
> high-profile users over to rawiso a.k.a. the current iso API of
> (lib)raw1394.
> 
> I suppose video1394 might get a viable migration path with your new
> driver, if you and interested developers put effort into development
> (and help with deployment) of a proper replacement.

As discussed on linux1394-devel, it may be possible to do a thin video1394 
compatibility driver for this one, but since the biggest user of this 
interface is libdc1394, it is probably better to just write a new iso 
streaming backend for this library. libdc1394 already supports different 
streaming backends.  For non-libdc1394 users of video1394, the interface I'm 
providing is very close to the video1394 ioctl interface, so porting should be 
easy enough.

>>  In the long run, supporting 4 different interfaces that does almost the same
>> thing isn't feasible.  The streaming interface in my new stack (only
>> transmission implemented at this point) can replace all of these interfaces.
> 
> You have to look at the matter not only from the POV of API design but
> also of deployment and support.

My POV here *is* about deployment and support, but from the kernel side of 
things.  If you commit yourself to long time support for the firewire stack, 
would you prefer 4 slightly different streaming drivers with different user 
space interfaces, or just one userspace driver with one userspace interface, 
that enables the 4 different types of streaming to be done in userspace?  The 
design of the streaming interfaces have been focused on enabling all these 
ad-hoc, in-kernel drivers to move to userspace, to make it feasible to 
actually support the stack.

Kristian