On Tue, 5 Mar 2002 09:57, Greg KH wrote:
> On Tue, Mar 05, 2002 at 09:52:47AM +1100, Brad Hards wrote:
> > On Tue, 5 Mar 2002 08:47, Greg KH wrote:
> > > On Tue, Mar 05, 2002 at 07:39:57AM +1100, Brad Hards wrote:
> > > > On Tue, 5 Mar 2002 06:37, Greg KH wrote:
> > > > > On Sun, Mar 03, 2002 at 02:33:20AM -0500, Matt Matthews wrote:
> > > > > > usbmouse                1824   0  (unused)
> > > > > > usbkbd                  2944   0  (unused)
> > > > >
> > > > > Can you just delete this modules to prevent them from getting
> > > > > loaded? You should not need them at all, and if both are loaded
> > > > > (along with the input drivers, not nice things have been known to
> > > > > happen.)
> > > >
> > > > I think maybe we should look at the Boot Protocol drivers again.
> > > >
> > > > 1. We seem to be confusing users, and we likely will get bit-rot on
> > > > the code 2. We now have dependency on the input drivers (ie. the
> > > > comments above probably aren't exactly true :). usbkbd no longer
> > > > calls handle_scancode directly - it gets passed through the input.o
> > > > and keybdev.o code. 3. The code size difference is there, but it
> > > > isn't that great.
> > > >
> > > > I'd like to suggest ripping out the boot protocol drivers, at least
> > > > in 2.5.
> > >
> > > No, they are still needed for some systems.  Think boot disks and other
> > > space constrained places.
> >
> > Is this a real requirement. Space isn't that constrained (bootable
> > business cards are taking over from floppies), and a quick check:
> > [bradh@localhost usb]$ ls -l usbkbd.o usbmouse.o hid.o
> > -rw-rw-r--    1 bradh    bradh       18258 Mar  5 09:27 hid.o
> > -rw-rw-r--    1 bradh    bradh        5928 Mar  5 09:27 usbkbd.o
> > -rw-rw-r--    1 bradh    bradh        4464 Mar  5 09:27 usbmouse.o
> > 8K isn't much for 99% of applications
>
> Tell that to the people who have to make distro boot disk images :)
They already deal with this problem using modules
> And remember, the bootable image on a CDROM still has the "old" 1.4Mb
> floppy size requirements.
But the driver doesn't have to be built in.

> > > > As a middle ground, we could just remove the config options, and
> > > > leave the code, along with a comment in Documentation.  I'd like to
> > > > do this for 2.4.
> > >
> > > No.  We currently have drivers in the tree that do different things,
> > > yet bind to the same type of devices.  Here's two pairs for example:
> > > bluetooth.c and hci_usb.c
> > >   ir-usb.c and irda-usb.c
> >
> > But usbkdb/usbmouse and hid don't do different things, so the analogy
> > doesn't hold. The examples you are presenting are presenting different
> > interfaces. hid is a superset of boot protocol, not a different
> > interface.
>
> Exactly.  But for some people, the usbkbd driver is very usable, and is
> all they really need.  The whole HID parser is overkill.  This is why
> the boot protocol was created in the first place.
Err, not so. Boot Protocol is meant to be used where you don't have a HID 
parser. We do. The whole usbkbd and usbmouse stuff is legacy we don't need.

> > A better analogy would be the uhci drivers. Do you want both of those?
>
> Ah, don't make me answer that one :)
>
> > > Should one of these pairs go away too?  I don't think so.  Each of them
> > > fill a need for someone, just like the Boot protocol drivers do.  It's
> > > up to the user to be "smart" enough to not enable the incorrect one.
> > > And it's up to us to document this to allow the user to make the
> > > correct "smart" decision :)
> >
> > Oh, how I have tried... :)
> > The depth of understanding required to really figure out what boot
> > protocol means is beyond most users. They aren't reading docs, and they
> > are stuffing it up.
>
> Then they fix it.  Remember, we aren't trying to make kernel
> configuration a task for Aunt Tillie here :)
They don't fix it. They come asking for help. There is a better way :)

> Sounds like a good FAQ question for the web page.
Maybe, but removing it is still better.

> > By contrast, the people trying to build systems that minimise RAM usage
> > and disk space will put the effort in.
>
> And we will continue to give them that option.  By "orphaning" the
> driver by removing the ability to easily build it, we subject the thing
> to potential bit-rot, and the accidental reintroduction of the option
> due to the main developers being forced to always enable it anyway for
> their testing (like me :)
So dump it completely, and stop stuffing around with it.

> > > I also remember some older USB keyboards that do not work with the full
> > > HID driver, but works fine with the boot protocol drivers.  I need to
> > > try to dig that device up.
> >
> > That is a different case. Now there is a downside to HID.
> > There are three options:
> > 1. Not support this hardware (relying on BIOS support)
>
> Nope, can't do that.  What if we _are_ the BIOS (think Linux BIOS).
Which could well use a decent HID parser.

> > 2. Keep the boot protocol driver, but modify it to match on vendor and
> > product IDs, since these aren't really HID devices (so blacklist them
> > like the wacom case)
>
> Nope, that breaks the USB spec.
Err, no. The device not working with full HID is in violation of the spec, us 
taking special action to deal with it isn't.

> > 3. Modify HID to deal with these devices (if possible).
>
> Much better.  I'll dig for my old device that only worked with one of
> the drivers later tonight and post if there is still a problem.
I like this option as well. Because then we can dump usbkbd and usbmouse :-)

Brad

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

Reply via email to