Re: NM is not being smart in a specific situation

[Dan Williams <dcbw redhat com>]

On Tue, 2006-09-05 at 23:35 -0300, Thiago Bauermann wrote:
> 2006/8/22, Robert Love <rml novell com>:
>         On Tue, 2006-08-22 at 15:39 -0400, Dan Williams wrote:
>         
>         > Though; we had discussed storing the mode of the network in
>         the
>         > info-daemon/GConf/Kconfig/whatever too, and ignoring access
>         points of
>         > that ESSID that don't match the mode.  That sounds
>         reasonable to me. 
>         
>         That sort of does the same thing as checking the BSSID, no?
>         
>         And we still have the same problem, if you see the _right_ AP
>         and try to
>         connect, but end up on the _wrong_ AP; there is always that
>         race.
> 
> Actually, the problem is that NM is selecting the wrong AP and thus
> feeding the wrong config info to wpa_supplicant. When I use
> wpa_supplicant alone with the right config, it associates with the
> right AP.
> 
> For some reason, NM is not restricting itself to the BSSIDs stored in
> gconf.

You mean that NM is not restricting itself to associating with APs it
sees in the scan list to the BSSIDs stored in GConf?  If so, that's
correct at this time.

We could autoselect the correct parameters based on BSSID, but we can't
tell wpa_supplicant to lock the association to a single BSSID, because
then we completely disable all roaming support between APs in the SSID.
I should think that storing the mode would be a better option here;
because obviously the mode should be the same.

Dan