2011/1/10 mike cloaked <mike.cloa...@gmail.com> > On Mon, Jan 10, 2011 at 4:19 PM, Franco Miceli > <fmic...@plan.ceibal.edu.uy> wrote: > > Mike, > > > > I have been dealing with the AP algorithm selection a few months ago. The > > one that selects the best connection is NM. If you get the src you can > see > > it in the file nm-device-wifi.c (for the wireless connections). > > > > The way it is implemented in nm 0.7 (which is the one I have been working > > on) is that NM will go through each favourite connection you have, and > > comparing them with the ones available from scan results. It does not > > prioritize on signal strength rather than timestamp. Which means if the > one > > you last connected to is available it will connect to that one. > > > > I have been playing with the code in order to make that selection signal > > sensitive. In order to do that you need to get signal strength info. The > > method nm_ap_get_strength(ap) can help get such data. The thing is that > not > > all wireless adapters report well signal strength, but if you trust > yours, > > then you can make this mod. > > > > Hope I could help. If you need I can share the code I used. > > Thanks Franco - I need to understand a few things - The version that > is current in F14 that I am running is > NetworkManager-0.8.1-10.git20100831.fc14.i686 and, as José Queiroz > mentions in his subsequent reply, there have been many important > changes since v7, although he does not reference a link to what those > changes have been in v0.8. > > However, what I do not know is whether the selection of AP was made > signal sensitive in the implemented changes in 0.8 or not? Certainly > the last signal used appears to be the one selected when booting up > the machine after switching off previously, which would be a > reasonable "first option" for the starting choice, but then a > secondary criterion where a stronger signal that already has been > used, and is listed in the connections in NM, would be selected and > switched in instead. Having such a two-stage selection would be what > would suit me (and many others) better than just going to the last one > used. > > I wonder if José Queiroz would tell us if this is in fact what has > been implemented in 0.8 (or 0.81 or 0.82) or not? That would be very > helpful. If this has been put into the code but has residual bugs > then presumably getting some data and putting in a bug report so that > developers could look at fixing it would be the way forward? > > A nice level headed and fact based approach to this would be ideal. >
I could swear that this was already discussed here, but the only reference I found was another message of mine. In the lack of good references, I ran to my kubuntu box. I also have an ESS in my home. My syslog shows several messages from wpa_supplicant about group rekeyings to one of the APs (the one that haves the strongest signal in the place I usually use my notebook). In one moment, I see messages from the kernel about deauthenticating the first AP, and authenticating the second one. Around these messages, I see only messages from wpa_supplicant, even though NM debugging is enabled. So I concluded that this reassociation was ordered by wpa_supplicant, not by NM. But, I admit, this is only an empirical verification; I didn't checked any code to confirm this.
_______________________________________________ networkmanager-list mailing list networkmanager-list@gnome.org http://mail.gnome.org/mailman/listinfo/networkmanager-list
