On Tue, Nov 10, 2015 at 11:52:12AM -0600, Andrew F. Davis wrote:

Anyway, All I'm trying to do here is keep things clean in the DT. We only have
one consistent option:

No, not really.

Match all sub parts by compatible:

Or we end up with some hybrid approach, matching some on node name, others
on compatible when needed. Yes, the above matches Linux device model (still
not sure why that is such a problem?), but it also matches modular functionality
in the device.

There's also the third option where we don't have any compatible strings
in the subnodes at all.