[PATCH RFC 0/3] Session support

Daniel Wagner daniel.wagner at bmw-carit.de
Fri Apr 9 08:16:54 PDT 2010


Hi Samuel,

On 09.04.2010 12:40, Samuel Ortiz wrote:
> On Thu, Apr 08, 2010 at 05:31:07PM +0200, Daniel Wagner wrote:
>> In order to get a better understanding on how that could be done.
>> I'm writing some proof of concept code (I'm allowed to release
>> that code under an open source license, if someone is interested).
> I would be interested, definitely.

It will take some more time to get the clearance for releasing it but
I expect this within couple of weeks.

>> During this work I already got some feeling for the problems. For
>> example assume with we have two 3G uplinks (OEM SIM and customer SIM)
>> available. Only application A is allowed to use both modems,
>> while application B only is allowed to use the second one. This is
>> exactly one of the scenarios I'm trying to implement. The current
>> API of connman is rich enough to get this job done without touching
>> connman.
> Please note that the session API is _not_ touching any of the exiting connman
> ones. It is an additional interface and applications are free to use it or
> ignore it.

Sure thing, I was trying to explain that all the things I want to do
are possible without the session API. So no problem.

>> Adding some sort of additional algorithm for choosing the "right"
>> service is not really helpful unless it can be completely
>> configured/adapted for a given use case.
>>
>> Do you plan to extend this interface?
> We currently don't because we have no requirements or use cases for that.
> That doesnt mean we're not willing to accept contributions from people coming
> with valid use cases for extending it. All the opposite in fact, as IVI is
> definitely a very interesting platform to us.

With the the code I'm writing I try to map map the (high level)
requirements to a technical solution. If it turns out that the code (or
algorithm) is okay to be merge into connman I'm glad to write some
patches (of course I need to get a okay to do so.)

>> Can I add my own fancy algorithm later through a plugin interface?
> That seems doable, yes. We should keep in mind that we want to keep the
> complexity away from the applications, only connman would know about those
> algorithms and which one is selected.

Sure thing. As I said it's exactly the thing I'm trying to solve for the
automotive area.

I could write some more about the things I'm trying to solve if this is
of interest.

cheers,
daniel




More information about the connman mailing list