Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.os.linux.networking > #4229

Re: networking a real dog's breakfast

Show all headers | View raw

William Unruh <unruh@invalid.ca> wrote:
> On 2018-08-29, Jerry Peters <jerry@example.invalid> wrote:
>> William Unruh <unruh@invalid.ca> wrote:
>>> On 2018-08-28, Jerry Peters <jerry@example.invalid> wrote:
>>>> William Unruh <unruh@invalid.ca> wrote:
>>>>> Networking on Linux seems to be a bit of a mess. 
>>>>> 
>>>>> wpa_supplicant just randomly connects with an access point, ignoring all
>>>>> suggestions as to what it should connect to. Thus, if you have (like me)
>>>>> a /etc/wpa_supplicant.conf file, it will just randomly choose one of
>>>>> them to connect to, if it is available. 
>>>>> 
>>>>> It will also just connect to the a given SSID in random order, rather
>>>>> than connecting to say the BSSID with the strongest signal available (
>>>>> as would be sensible) Thus it will connect to a flakey 30% signal
>>>>> strength BSSID rather than a 90% signal strength BSSID with the same
>>>>> SSID.
>>>>> 
>>>>> iwlist craps out if there are too many BSSID available. When I am at
>>>>> airports, it is impossible to use the Network Center on Mageia, since it
>>>>> uses iwlist, and airports often have "too many" access points for iwlist
>>>>> (iw will list them, but it has a totally lunatic output format, and
>>>>> gives huge warnings that you should not screen scrape it-- and it has,
>>>>> as far as I can see, no other output.) Also iw seems not to work at all
>>>>> half the time, or if you happen to be connected to an access point, it
>>>>> will only give that one AP, when there another 50 available that it does
>>>>> not list. 
>>>>> wpa_cli MUST have wpa_supplicant running when it tries to do anything. 
>>>>> 
>>>>> And all of them have woefully inadequate documentation, and ancient
>>>>> documentation. ( for example, wpa_supplicant documentation does not mention
>>>>> nl80211 driver, when that is now the strongly approved driver, all
>>>>> others being deprecated. )
>>>>
>>>> Use networkmanager it will control wpa-supplicant & internally it uses
>>>> the nl80211 interface. Wicd on the other hand, calls iwlist internally
>>>> so in your case it'd be problematic.
>>>>
>>>> You also will need nm-applet to convienently set up each network.
>>> 
>>> I did try the network manager about 3 years ago for a while, and found
>>> it frustrating. When it worked it was fine. When it did not it was very
>>> hard to debug or to figure out how to fix it. network center (on Mageia)
>>> is much more configurable, but the person who wrote it is long gone, and
>>> the others seem to be leary/uncomforable at tackling it. It is not a
>>> well documented code at all (in perl), and is somewhat obscure. I think
>>> it eventually uses the ifup scripts to do the actual connecting. One
>>> problem is that wpa_supplicant itself is weird. For example it does not
>>> connect to the strongest signal for an SSID, but goes through trying the
>>> first one that reported, then if that does not work, the next, etc. Thus
>>> with one AP having strength 95% and another with 44%, it is as likely as
>>> not to connect to the weaker one, making the connection very flakey.
>>
>> You might experiment with NM again, it's under continual development
>> and there are significant changes between versions. One problem that I
>> have with NM is that it's typical GNOME -- it assumes that its way of
>> doing things is what everyone wants.  I want dnsmasq to be a caching
>> nameserver, and don't care about "split dns for vpn", so I use a
>> wrapper script to ignore most of the args NM passes to dnsmasq so I
>> can use it the way I want. 
>>
>> Wicd is simpler, but again uses iwlist internally and doesn't seem to
>> have a large developer base. NM also connects faster than wicd,
>> probably related to using nl80211 directly rather than parsing iwlist
>> output.
> 
> If they use iw to read the APs then it is far more difficult to parse
> than is iwlist. (iwlist unfortuantely has a max buffer size too small
> for man locations). However, NM might use wpa_cli to read from
> wpa_supplicant the APs that are available. Not sure how to run
> wpa_supplicant so that it does not try to connect to anything but does
> actually scan all the APs, so wpa_cli can read them and then tell
> wpa_supplicant to connect to a certain one. 

AFAIK NM uses nl80211 directly via netlink, not an external program
like iw.
There's also an option in wicd to use another python module instead of
iwlist, but I have no idea what interface that module uses.

> 
> Also, if you have many BSSIDs which have the same SSID name, does NM
> connect to the strongest one, or does it also randomly choose (well in
> order of actually receiving the scan, which is random with respect to
> the strength)?

Sorry, don't know, you'd have to experiment to find out, unless it's
in the NM documentation.

I just switched from wicd about 2 months ago when a kernel upgade
broke iwlist for ath5k (now fixed, btw). Once I got used to NM I find
that I like it better than wicd. A big reason is that it automatically
reconnects after resume from hibernate or sleep, something wicd would
do sporadically. 

Back to comp.os.linux.networking | Previous | Next — Previous in thread | Find similar | Unroll thread

Thread

networking a real dog's breakfast William Unruh <unruh@invalid.ca> - 2018-08-20 19:34 +0000
  Re: networking a real dog's breakfast Jorgen Grahn <grahn+nntp@snipabacken.se> - 2018-08-20 20:12 +0000
    Re: networking a real dog's breakfast William Unruh <unruh@invalid.ca> - 2018-08-20 20:59 +0000
    Re: networking a real dog's breakfast tom <tom@0.0.0.0> - 2018-08-27 21:19 -0700
      Re: networking a real dog's breakfast Dan Purgert <dan@djph.net> - 2018-08-28 11:29 +0000
        Re: networking a real dog's breakfast tom <tom@0.0.0.0> - 2018-11-25 18:43 -0800
      Re: networking a real dog's breakfast Allodoxaphobia <knock_yourself_out@example.net> - 2018-08-28 13:31 +0000
  Re: networking a real dog's breakfast Jerry Peters <jerry@example.invalid> - 2018-08-28 20:10 +0000
    Re: networking a real dog's breakfast William Unruh <unruh@invalid.ca> - 2018-08-28 20:49 +0000
      Re: networking a real dog's breakfast Jerry Peters <jerry@example.invalid> - 2018-08-29 20:21 +0000
        Re: networking a real dog's breakfast William Unruh <unruh@invalid.ca> - 2018-08-29 23:29 +0000
          Re: networking a real dog's breakfast Jerry Peters <jerry@example.invalid> - 2018-08-30 20:19 +0000

csiph-web