Re: [RFC PATCH 3/4] PLATFORM: Introduce async platform_data attach api

[Greg KH <greg@kroah.com>]

On Mon, Mar 14, 2011 at 05:03:17PM -0400, Alan Stern wrote:
> On Mon, 14 Mar 2011, Greg KH wrote:
> 
> > No, it has been determined a long time ago that network naming things
> > like this are to be done in userspace.  It's an argument that has come
> > and gone many years ago, sorry.  See all of the wonderful, and simple,
> > tools we have today in userspace to handle this type of thing.  Distros
> > can use them how ever they see fit, and even better, users can configure
> > them!  That means they don't have to rebuild their kernels, which is a
> > bit unreasonable, don't you think?
> 
> ...
> 
> > Perhaps we should just always name these things 'eth%d'?  Oh wait, as it
> > really is a USB device, they are supposed to be called 'usb%d' as
> > determined (again) a long time ago.
> > 
> > If a distro/board manufacturer wants to hide the fact that this really
> > is a usb device by renaming it to eth0, then again, it can.  But don't
> > force the kernel to have that policy in it.
> 
> This argument does sound contradictory.  If network interface naming 
> should be left entirely up to userspace, then why doesn't the kernel 
> always generate names of the form "eth%d"?  Why not rip all that stuff 
> about "usb%d" or "wlan%d" out of the driver entirely?
> 
> (Apart from the fact that this would be a user-visible change in kernel 
> policy and would break a large number of systems...)

I think that is the only reason it is sticking around.

thanks,

greg k-h