Re: Ad-hoc vs user-created

[Dan Williams <dcbw redhat com>]

On Tue, 2006-06-20 at 14:32 +0800, Bernard Blackham wrote:
> On Mon, 2006-06-19 at 22:14 -0400, Dan Williams wrote:
> > > As user-created implies ad-hoc, but not vice-versa, simply testing
> > > for ad-hoc instead of user-created in supplicant_send_network_config
> > > fixes my issues. (Specifically, setting ap_scan = "AP_SCAN 2", and
> > > setting ad-hoc mode). Patch attached (applies against current CVS
> > > too).
> > 
> > Right; user_created is a subset of Ad-Hoc.  We have to specialize
> > user_created because we're not running a DHCP _server_ locally, so
> > user_created networks will do auto-ip/link-local addressing.  Ad-Hoc
> > networks won't generally do that, although we do fall back to it if DHCP
> > fails for adhoc.
> 
> I guess that if you "create" an ad-hoc network even though it already
> exists (as opposed to joining it), it'll skip straight to link-local
> too?

Correct; but it's unclear why you'd do that.  If you see the network in
your scan list, then you should just be able to click on it and NM will
connect to that adhoc network and try to get DHCP.  If DHCP times out,
it will do autoip.

Dan

> > > Though if I've done something grossly wrong, feel free to flame me
> > > instead :)
> > 
> > Nope, patch looks good, other than a few things; first, we get to rename
> > stuff more extensively.  The user_created arguments for the
> > nm-ap-security* stuff should be changed to adhoc, and where we pass
> > user_created into the supplicant config function, we really do want
> > adhoc there instead since you can't use some WPA options with adhoc.
> > 
> > So I've cleaned up these small bits and will commit to HEAD and STABLE,
> > thanks.
> 
> Great, thanks!
> 
> Kind regards,
> 
> Bernard.
>