On Tuesday 15 January 2008 20:38, Carlos Corbacho wrote:

On Tuesday 15 January 2008 20:54:29 Dmitry Torokhov wrote:

quoted

If there is a [planned] WMI sysfs device then I think input device should
use it to form proper sysfs hierarchy. What are the roadblocks for getting
WMI sysfs in?

The question is, would exporting the GUID virtual device and using that as the 
parent make more sense? So, in the case of hp-wmi, it only uses the GUID 
95F24279-4D7B-4334-9387-ACCDC67EF61C, so should we use the virtual device 
associated with that GUID as the parent, or just have WMI itself as the 
parent device? I'm really not sure which would be the better way to go.

Why not used generic names like dev0, dev1 etc.
and have the guid be an attribute?

Unfortunately, as you can see, a GUID is a 36 character string, which is 
longer than the current 20 byte length of bus_id in a 'struct device'; so 
either we have to have a temporary hack to shorten the GUID length when 
creating the device, or wait for bus_id to switch to a variable length 
(although it's not certain if this change will hit 2.6.25, or will be put off 
until 2.6.26).

-Carlos