Re: Some various thoughts on PPP

[Dan Williams <dcbw redhat com>]

On Fri, 2007-05-18 at 14:27 +0300, Tambet Ingo wrote:
> On 5/18/07, Ulrik Mikaelsson <ulrik mikaelsson gmail com> wrote:
> > One of the things that eludes me though; how exactly are you
> > envisioning the NM-dialing-connection process? From what I understand,
> > you're planning to have NM itself dial, and then hand-over the
> > connection to pppd, linked in to NM? Will this be carried out in a
> > separate process, with DBUS-connection to the main process? Is this
> > somehow supporting the desire to keep network-configuration
> > centralized within the distribution, or will it require configuration
> > separate from, say /etc/ppp/peers?
> 
> NM uses a new device NMDeviceModem to dial out (in process) and starts
> pppd process. The dbus interface will remain the same as for all the
> other devices, just the activation structure will contain modem and
> ppp specific data. If we want (and probably do), we can add a new dbus
> interface for dialup devices (just like wired and wireless devices
> have their own device specific interface in addition to the standard
> device interface) that has additional status signals for better
> progress reporting (like, "initializing modem", "dialing", ...).
> pppd process will emit dbus signals (with help from a pppd plugin) so
> NM knows what's going on.
> 
> > > That is yet not all, it also turns out HAL doesn't have any notion of
> > > modems so the best we can do for now is to get a list of HAL devices
> > > with capability "serial" and try to send some modem initialization
> > > strings and see if they respond. That is for the serial devices that
> > > actually show up in HAL at all. Most of the thinkpads have a modem
> > > that only works with proprietary HSF modem driver and these do not
> > > show up in HAL at all.
> > Well, perhaps there's another quirk to it, as far as I understand,
> > many modems expose several devices, for different uses. Especially,
> > the GRPS-datacard I tested had three of them, one for the
> > ppp-connection, one for chatting with the card simultaneously with the
> > connection, querying signal strength and such, and one for to me
> > unknown purposes. Somehow we'll gotta figure which is which. :)
> 
> Yeah, this is unfortunate. Pretty much identical to the wireless
> situation until wpa_supplicant provided us with a standard interface.
> I guess for each of these multi-device card we'll have to write a
> NMDeviceModem subclass, ugh.

Most broadband cards I've seen have the first serial interface be the
actual data interface.  But that's not really helpful.  We may have to
create HAL fdi files to tell HAL about the different broadband cards and
phones and such.  But I think this can live in HAL, and we can just have
the .fdi files assign a property to the device to tag it as one we
can/should use, including the serial interface we care about.

Dan