Re: [PATCH] Delay full modem initialization until SIM is unlocked

[Dennis Noordsij <dennis noordsij helsinki fi>]

Tambet Ingo schreef:
>
> This is actually needed for a lot of functionality, but it looks like
> the monitoring device is not standardized and (almost?) every device
> has it's own way of doing things. Some have proprietary binary
> formats, some have AT commands. If it's done in NM, I'd like to
> implement a plugin system for this, otherwise the code will get
> unreadable and unmaintainable very quickly...
>
> Another possibility we've been considering is to delegate all the
> dialup handling to an out of process helper program, much like
> wpa_supplicant is used for wireless. The lack of standards and lack of
> implementing existing standards in mobile devices makes it a very
> difficult job to support a wide range of hardware and it probably
> deserves a dedicated program. The requirements from NM are that the
> program implements easy to use high level dbus interfaces that
> abstract all the quirks (again, much like wpa_supplicant). Currently,
> the only candidate for this is VMC, Pablo could give more information
> on that.
>
>   

Thanks for your quick reply!

The seperate helper program is actually a nice idea IMO, makes NM more
about the logic
of managing the devices, and lets the device complexities be handled by
the external
helper program. One can then really take advantage of device specific
quirks and features,
and more easily fix/improve things without possibly disrupting support
for other devices.

And one could more easily contribute without having to touch the NM
core, and make
everything language agnostic.

What kind of dbus API would you propose ? What would be the workload in
NM to
adapt to use such a dbus interface ? Any additional thoughts/ideas you
have about this ?

It would perhaps be fastest to proceed like:
- define an API which is perfect from NM's point of view
- Make the GSM (and CDMA) devices seperate helpers (i.e. maintain
current functionality)
- A helper dispatch mechanism to use the correct helper based on the HAL
data (with a fallback
on the current code if nothing more specific is found)

At this point people can experiment with custom helpers (possibly based
on VMC) however they wish.

I'd like to try something like this for my device.

Any thoughts ?

Dennis