[Kde-hardware-devel] Wifi interface

[Stefan Winter]

Hi,

> That's definitely something that could be addressed in the backends (the
> NetworkManager backend for example). As you probably noticed the interfaces
> exposed through Solid are too abstract for this.

Yep, see my other mail from a few seconds ago.

> > Oh, and my earlier comment still is valid in this regard:
> > QStringList ...addresses, instead of QString ...address;
>
> Ok, I just put an extra comment regarding this one because I admit I'm a
> bit split on this issue. I tend to think that for the average desktop usage
> you have only one address on the network not several of them. That looks
> more like a setup used for servers, so the questions are:
> 1) Is such a setup common on desktop systems?
> 2) Do we want to support a feature like this one more server oriented?

1) IPv4: No. More than one address is _quite_ exotic. "QString address" would 
do if you so like.
    IPv6: Yes. Two addresses is the bare minimum when you are connected to a 
native IPv6 network. Addr 1 is the so-called "link-local" address and begins 
with fe80::, is only valid on a per-interface basis (that's where the scope 
id's from my earlier mail come in) and is not being routed; Addr 2 is your 
globally valid, routed IPv6 address. So in fact, a "QString address" will 
NEVER be adequate if you are using IPv6. Here, a "QStringList addresses" is 
must.
More than two are also common: typically, you get a computed globally unique 
IPv6 address in your subnet that's computed out of your MAC address. You may 
not like that because not a "beautiful" address. So you add a "nice" one. 
Then you have three: fe80::, coomputed_addr, manually_configured_addr. I'll 
send an ifconfig output of my work workstation tomorrow to illustrate this.

2) As a corrolary of 1): for IPv4 support is optional and probably a waste of 
time. IPv6: We _have to_ support this.

> I agree, hence why the enum got extended (Unassociated is back, Master and
> Repeater are now in).

Great!

> Added a comment for this one. It can probably get simplified a lot (or
> simply disappear) thanks to the Encryption classes. It also addresses
> Christopher's comment regarding the fact that it's pretty much tied to
> NetworkManager, and I'd dislike to have something that backend specific in
> the libs.

We should just keep it in mind. I don't claim to have the one-and-only answer, 
just wanted to point out possible points of friction.

> Well, I think you missed the point here (hence why I ignored your comment).
> It's here to allow us to determine that two networks with the same ESSID
> but different APs are probably different (think about the "netgear" SSID).
> Two networks (same ssid, different aps) could then be merged if we know
> (for example we asked the user) that they're really the same network.

See the other mail, the description confused me. But forget about asking the 
user. What would you answer if a popup asked you:
"There are two APs that are on SSID netgear: 00:13:24:22:11:a1 , 
00:27:fe:2a:e1:aa. Are you sure these two belong to the same network?"
(Yes) (No) (WTF?)

My answer (if it wasn't my home network which I set up and administer myself 
and so by chance might know the MAC addresses) would definitely be "WTF?"

The "if we know" is actually the very very tricky part. You can only find out 
by connecting and pinging layer 3.

> > +    signals:
> > +        virtual void signalStrengthChanged( int );
>
> Yep and that's fine IMHO.

Okay.

> > I didn't see outOfRange() in the diff. If it's still in the code, see my
> > remarks on its reliability.
>
> No it's not there anymore.

Okay.

> > I expect my cookies to be delivered at aKademy :-)
>
> Sure! <mental_note>Bring cookies for Stefan.</mental_note>
> That said you know that it'll probably be easier to find beer than cookies
> in Dublin. ;-)

I might settle for beer as well. :-)

> Looking forward to meet you there.
>
> OT: You weren't at previous aKademies, were you?

I was in Malaga. We must have missed us. I mostly collected reprimands there 
for the freaky kcmwifi that I did when I was young and stupid ;-) .

Greetings,

Stefan

-- 
This mail is guaranteed to be virus free because it was sent from a computer 
running Linux.

-- 
The K Desktop Environment
- Stefan Winter -
Areas of Activity:
kdenetwork/wifi (KWiFiManager)
kde-i18n/de (German translation)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 191 bytes
Desc: not available
Url : http://mail.kde.org/pipermail/kde-hardware-devel/attachments/20060809/ea22d3ef/attachment.pgp