> > Why is "i2c-alias-pool" in the drivers binding and not a regular i2c
> > binding? Same question for the implementation of the alias-pool
> > handling. Shouldn't this be in the i2c-atr library? I'd think managing
> > the list of aliases would look all the same in the drivers otherwise?
> 
> I think that this _was_ the plan, as it looks obviously cleaner, but
> then we agreed that we should remove the pool entirely, so I didn't
> bother moving it.

Ah, you mean we agreed on that at the Plumbers BoF? I think we can
conclude this is obsolete meanwhile. GMSL encodes the target addresses
in DT. Rob is also fine with the binding here to encode the pool in DT.
Let's follow that road, I'd say.