Re: [PULL] http://www.linuxtv.org/hg/~hverkuil/v4l-dvb-ng

[Laurent Pinchart <laurent.pinchart@skynet.be>]

Hi Hans,

On Monday 01 December 2008, Hans Verkuil wrote:
> On Monday 01 December 2008 15:24:43 Laurent Pinchart wrote:
[snip]
> > In a few months time (probably even earlier) the v4l2_device
> > structure will be reworked (and possible renamed). I'm fine with it
> > going to linux-next now if we agree on the following.
> >
> > - We should only advocate v4l2_device usage for subdevices-aware
> > video devices. Porting all drivers to v4l2_device is currently
> > pointless and will only make future transitions more difficult.
>
> Agreed. For now it is only relevant for drivers that use subdevices.
>
> > - v4l2_device should be marked as experimental. I don't want to hear
> > any API/ABI breakage argument in a few months time when the framework
> > will evolve.
>
> Am I overlooking something? This API is a kernel API, not a public API.
> Hence if I (or anyone else for that matter) make future changes then it
> is my responsibility to adapt all other drivers that are affected at
> the same time. I don't see how any of this could break compatibility.
> Except for out-of-kernel drivers, of course. But that's the risk that
> they always run.

You're right. It might be useful to state that the API is a work in progress 
in the documentation, but I'll let you decide on that.

> Marking this API as experimental seems pointless to me. It either works
> and so is available for use or it doesn't and then it is a plain old
> bug that needs to be fixed. I also know already that there will be
> changes as e.g. sensors require a new ops category and v4l2_device
> might need a notifier callback as well. However, I'm not going to
> implement that until there is also a driver that uses it (adding
> functionality to an internal API just because it might be needed in the
> future is a really bad idea).

Best regards,

Laurent Pinchart