Re: proposal for deletion of drivers/hid/hid-ids.h

[Oliver Neukum <oneukum@suse.de>]

On Thu, 2015-03-26 at 10:06 -0400, Benjamin Tissoires wrote:
> On Thu, Mar 26, 2015 at 7:44 AM, Oliver Neukum <oneukum@suse.de> wrote:
> > Hi,
> >
> > I would like to kill drivers/hid/hid-ids.h and replace it
> > with numerical IDs in the files using it.
> >
> > There are two reasons for that.
> >
> > 1. It is a layering violation. There should not be a private
> > data base for USB IDs in HID.
> 
> Technically, this DB is not only for USB devices. We also have
> Bluetooth and I2C devices here.

Well, the token IDs ;)

> > 2. It serves no purpose and adds work. Anyone who adds a quirk
> > or a special case for devices needs to operate on the numbers,
> > as those are what he gets from the descriptors. Looking up or
> > adding a symbolic name for a device is just more work without
> > a benefit. These numbers have no intrinsic meaning beyond
> > being unique and it rarely matters (and should not matter)
> > for which vendor a particular fix is intended.
> 
> I disagree. This list might not be useful for the
> drivers/hid/usbhid/hid-quirks.c by itself in most cases.
> However, we mainly rely on this list to add the device in
> hid_have_special_driver and hid_ignore in hid-core and in the
> submodule that should handle it.

Can you explain how we depend on it? We certainly use it, but
how do we depend on it? I don't see how just the numbers would
be worse. In fact they would be better as you again save a lookup.

> Many times, already having the VID/PID registered in hid-ids.h saved
> some time when debugging and adding a subdriver for a special device
> because if the VID/PID is already in hid-ids.h, that means that

Again, I see how having the VID/PID pair is an advantage. I don't
see why having symbolic names for that pair is an advantage.
Just having the numbers in a list of quirky devices would serve
the same purpose.

> someone already dealt with it, and it gives us a way to clean it when
> the quirk was not appropriate. For instance, many multitouch devices
> were added before the creation of hid-multitouch and were registered
> with the quirk MULTI_INPUT.

Well, yes, so you needed to grep for MULTI_INPUT. The entries would
still have been present, just with nummerical IDs.

	Regards
		Oliver