Re: Handling GetSecrets with request_new == true

[Dan Williams <dcbw redhat com>]

On Fri, 2009-11-27 at 00:57 +0100, Will Stephenson wrote:
> I often observe password dialogs caused by a GetSecrets with request_new == 
> true when the secrets are correct, for example due to unfavourable wireless 
> propagation conditions.  This blocks the machine from reconnecting until a 
> user OKs the dialog.  
> 
> Would it be very wrong to automatically Ok the dialog after a countdown, so 
> temporary disconnections can resolve themselves?

Possibly not; but something that's been on my list for a looooong time
has been to  make NM just try harder and only request new secrets as a
last resort.  Like if there's only one AP in the area that NM could
possibly connect to, and it fails, try again.  If that fails, wait a bit
and try again.  If that fails, wait a bit longer and try again.  Same
for ethernet.

Basically, the idea originally was that your password was wrong and you
should re-enter it.  That was valid in 2005, but is no longer really
valid.  These days the password is usually correct and the driver or
connection is just shitty.  If the password actually isnt' correct,
there are better ways of handling this, and it's something we need to
enhance wpa_supplicant for so we can actually propagate the "your
password is wrong" error back up to the user in a meaningful way.

Dan