Spec suggestions regarding IDE

[Artem Kachitchkine]

> I think the ide_host bus should be removed from the spec in lieu of a 
> general ide.host capability.

ide_host and ide namespaces represent different physical entities.

ide_host.* describe an IDE controller.
ide.* describe an IDE device, connected to an IDE controller.

ide_host typically has a parent node, usually of the 'pci' bus. This pci 
node also describes the IDE controller, but from the PCI standpoint. 
Since IDE controllers are essentially peripheral bus to IDE bridges, the 
two nodes reflect this duality. So you can have:

pci
   ide_host
     ide
       storage

> Currently the definition 
> of an IDE host according to HAL (afaics) is an IDE device which hosts 
> two devices, primary/slave.

I'm not sure I understand. In the above hierarchy, you can have have as 
many child nodes as necessary, on each level. If you have two IDE 
controllers, with master and slave on the first, and only master on the 
second, the tree will look like:

pci
   ide_host
     ide
       storage
     ide
       storage
pci
   ide_host
     ide
       storage

Or am I missing something?

-Artem.