Re: [PATCH] make psaux reconnect adjustable

[Gunther Mayer]

Kurt Garloff wrote:
> 
> On Tue, Aug 14, 2001 at 09:58:55AM -0700, Linus Torvalds wrote:
> >  - do we actually need the config switch AT ALL, whether at bootup or not?
> >    What exactly breaks if we just always pass the AA 00 values through?
> >    Apparently nothing ever breaks, which makes me suspect that people are
> >    just being unnecessarily defensive.
> >
> > In short, I'd prefer a patch that just unconditionally removes the code,
> > unless somebody KNOWS that it could break something. That failing, a
> > simple kernel command line option sounds better than more files in /proc.

Linus, no more boot options (and no more files in /proc), please.

Don't push policy on users when there is a perfect, simple and _user-friendly_ solution.

> 
> OK, here come two patches. The first one removes the special PSAUX reconnect
> handling completely. So userspace should handle it. (Which is possible; just
> not implemented in gpm/X11 at this time AFAIK.)
> 
> Second patch reintroduces the special handling again, but does
> * react on AA 00 instead of just AA, thus much less likely breaking other
>   drivers (such as synps2). All PS/2 mouses I could access (about 5
>   different models) produced AA 00, so this seems OK.
> * is disabled by default, and needs to be enabled by the psaux-reconnect
>   boot parameter, like in 2.2.19.

Your patch for "AA 00" is preferable over current state "AA" (<=2.4.8)

ioctl is preferable over sysctl. Aware drivers do "ioctl("/dev/psaux", PS2_TRANSPARENT)".

run-time configuration is preferable over a boot parameter (see above about user-friendliness).

Not breaking current behaviour ("reconnecting 3-byte protocol mice works perfectly") is preferable.