On Thu, Nov 26, 2009 at 06:40:04PM +0000, Jonathan Cameron wrote:
> Greg KH wrote:
> > On Thu, Nov 26, 2009 at 05:19:13PM +0000, Jonathan Cameron wrote:
> >>> That being said... If we want user-space to know what device is there,
> >>> we may want to still let drivers pass a name string to
> >>> als_device_register() and let the ALS core create a "name" sysfs
> >>> attribute returning the string in question. This would be much lighter
> >>> (for individual drivers) than the previous situation, as the string in
> >>> question would be a constant (e.g. "TSL2550".) Opinions?
> >>>
> >> Makes sense given we want all drivers to support some form of identification.
> >> We could do it by stating they will all have that attribute, but given it's constant
> >> will save repetition to put it in the driver. Conversely it might complicate the handling
> >> of subsequent attribute_groups so I'd probably favour adding relevant documentation lines
> >> and leaving it up to the drivers to implement this attribute.
> >>
> >> Thus we'd require (within reason) all drivers to have illuminance0 and name.
> > 
> > Why have a name attribute when you can just use the name of the device
> > itself instead?  Isn't that what it is there for?
> Could do, though I'm not entirely sure all bus types are implementing a name
> attribute (I may be wrong, but I don't think spi does for example though it might
> have gone in with the recent device table stuff).  We could just specify that it
> should be present for the device.

ah, sorry, I was thinking of the name of the actual device, which is the
bus id here.

Nevermind, I'll go back to feeling stupid...

greg k-h