On Sat, Apr 27, 2002 at 12:12:54AM +0200, Oliver Neukum wrote:
> 
> What I do question is whether it is worth the trouble.
> You get a lot of trouble, like a completly new, unproven,
> supporting infrastructure for relatively little gain.

The people with 20 printers or scanners have something to gain (think
testing labs) :)

> Yes, you are right. I thought about this this morning.
> To me the obvious solution would have been to use devfs
> and request a dynamic _major_ for every driver that implements
> a char device.
> In fact would you like an API change to move the devfs stuff into
> usbcore ? I see no reason every driver registers itself with devfs.

I hate the current implementation of devfs support in the USB drivers
too.  I'd be willing to take such a patch.

> > Please explain.  How could this be done within today's framework?  And
> > remember, 99.9% of all machines do not use devfs today :)  This patch
> > solves a real problem for all machines, without requiring people to use
> > devfs.
> 
> It requires them to use something new totally unproven.
> I somehow fail to see the advantage ;-)

The scheme is proven in the usb-serial drivers which use the same kind
of minor number sharing.  I haven't heard of any complaints about how
that is implemented :)

thanks,

greg k-h

_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to