Proposal v2: identifying modems and mobile broadband cards

[Dan Williams]

On Mon, 2008-02-11 at 21:05 +0200, Faidon Liambotis wrote:
> Dan Williams wrote:
> > - An interface with the 'modem' capability may also acquire the property
> > 'modem.at_command_sets', which is a string array of supported command
> > sets, denoted by their abbreviated standards name. If the property is
> > present, it is assumed the modem supports rudimentary Hayes-compatible
> > AT commands.  The key may contain _one or more_ of the following values:
> > 
> >    a) IS-707-A  (for CDMA supporting cards)
> >    b) GSM-07-07  (for GSM supporting cards)
> >    c) GSM-07-05  (for GSM supporting cards)
> There are some GSM/UMTS/HSDPA cards out there that are based on some 
> Qualcomm chipset (namely, the Novatel ones, including the ones found on 
> recent Dell laptops) that support the GSM command set on ttyUSB0 but a 
> completely different binary protocol on the management port, ttyUSB1.

Right; the problem here is that when you've got an active data
connection on ttyUSB0, you _have_ to communicate with the card via the
proprietary interface on ttyUSB1 to do things liek signal strength and
whatnot.

> This binary protocol is called BREW[1]. On the Novatel card that I own, 
> it is possible to pass-through the standard GSM AT command set via BREW 
> and I've made some success towards this point.

Great to hear.

> In the event that I manage to complete the reverse engineering (I'm not 
> very far from it), it'd be nice if HAL would transparently support 
> talking to the management port. Not sure how, but it seems like a good 
> time to include this on your plan.

Yeah; if we use 'modem.command_sets', we can add 'brew' as a supported
command set there.

> BREW is apparently used on some CDMA cards but I have no idea if the 
> same AT command passthrough works there.
> May be this should be yet another key for those CDMA cards, not sure if 
> (a) includes it.

Good question; I believe that all current CDMA products need to support
IS-707-A on the primary serial interface, but if Qualcomm provides
equivalent functionality with BREW then maybe the vendors don't all
support AT passthrough on the other interfaces.

> Note that there is already a free software implementation for a part of 
> BREW, called BitPim[2]. This software is for CDMA phone management 
> (phonebook, SMS etc.). Unfortunately, there are no indications that they 
> are even aware of the AT passthrough feature.

Yeah, heard of bitpim; haven't looked too closely at it though.  Do you
have a site where you are developing a specification for BREW?

Dan