Re: [PATCH 5/7] Make the driver use MSC_SCAN and a setkeycode and getkeycode key table.

[Matthew Garrett <mjg59@srcf.ucam.org>]

On Tue, Jul 17, 2007 at 11:37:33AM +0100, Richard Hughes wrote:
> On Mon, 2007-07-16 at 19:09 +0100, Matthew Garrett wrote:
> > 
> > That's fine, we have information that lets us distinguish between 
> > models. 
> 
> Yes, but the consensus I seemed to get from previous mails was that the
> matching quirks should be done in userspace where possible. The current
> FDI thing lets a user edit a simple text xml file and stuff just works.
> I'm not sure asking user to compile custom kernel modules to test out
> new keymaps is feasible, at least for the majority of linux users.

That doesn't match what the kernel currently does. It's still possible 
to send the custom scancode (unless I'm entirely misreading this) and 
remap if the user genuinely does have a weird keymap, so all you're 
doing is moving small tables from one part of disk to the other - 
they'll be flagged __initdata, so there's no runtime overhead. On the 
other hand, you get the advantage that the keys just work.

-- 
Matthew Garrett | mjg59@srcf.ucam.org