On Jan 12, 2007, at 4:43 AM, Sylvain Munaut wrote:

Sascha Hauer wrote:

quoted

On Fri, Jan 12, 2007 at 10:00:31AM +0100, Sylvain Munaut wrote:

quoted

I'm not saying the system is perfect but this issue is more  
related to
recent
bindings that the of thing as a whole. That's more the problem  
with of
is that
when you need to define bindings for something that has none, you  
may not
anticipate everything ....

No, you can never do that. Lets hope there will never be incompatible
changes in the device tree so that we have to use this device tree  
for this
kernel version and another one for other kernel versions.

I would *not* base anything (production) on the current bindings,  
because
the changes we consider now are pretty "basic" and not compatible.

But it's clear that it's probably the last opportunity we have to do
such changes,
afterwards we'll be bound to whatever we decide in the next few weeks.

Do realize you can use the device tree in a much simpler form to pass  
the same basic information we got from the bd_t and ignore everything  
else.

The reason for all the churn (or complexity) is the lack of any  
'spec' on how to describe any of the SOC parts in the device-tree and  
thus we have to invent something.

At least that's how I see it.

- k