Hi Marcel,

On 10/18/2011 06:32 PM, Marcel Holtmann wrote:
>>>>> To be honest this all sounds fine on the surface, but once I look at the
>>>>> implications I am having problems to come up with a good reason on doing
>>>>> this. I am currently not seeing this need.
>>>>
>>>> If an application knows that the data link is suspended it could stop
>>>> all timeout timers because it doesn't make sense to keep them running.
>>>>
>>>> If the link resumes the timer can also be resumed. This is what I have
>>>> heard is working quite good for our current implementation here and
>>>> hence the request to have something similar around.
>>>>
>>>> I see that oFono exports this information through the Suspended property
>>>> in the ConnectionManager API. I didn't see anything like this on the
>>>> BlueZ APIs (when ConnMan uses PAN). So this information is already there
>>>> and can be used at least for real modems. I don't really understand your
>>>> argument about the implications.
>>>
>>> The oFono Suspended property is different. That is a GPRS network
>>> specific details of suspended data connection when you get a phone call.
>>> It is a limitation of being on a GSM network. And that one is just
>>> present so the UI can tell the user that its connection got suspended.
>>
>> This is what I was trying to explain what I want to signal to the
>> application.
>>
>>> Keep in mind that this is also only useful for a limited amount time
>>> where the network holds your IP details and your TCP connection timeout
>>> did not kick in. If we are talking of 1 hour phone call you are just out
>>> of luck anyway.
>>
>> Sure, for short phone calls it should work although.
> 
> So we could signal this. I was just more thinking of just signaling that
> we lost the IP address. And give it back later.

The application could distinguish between temporary suspended and a real
connection lost. Well, it is hard to tell either way.

> This would make the logic in the application simpler, but it also means
> that we can not survive small suspend windows. Or we have to do some
> internal magic in keeping the IP address for some time before we expire
> it.

Not sure if you really have to make it that complicated for this. For
those application I have in mind we have to read the oFono API's anyway
(MCC, MNC, Signal Strength). Reading an additional value is easy.

>>> So far enough, GSM networks can signal a suspended state while
>>> potentially keeping the IP address (TCP connection might still timeout).
>>>
>>> Where else do you see this?
>>
>> I am not sure if this kind of information would accessible over
>> Bluetooth if we have a smartphone connected to the head unit and we are
>> doing HFP and PAN at the same time. Need to dig into the standards first.
> 
> The suspend only happens on GSM. With UMTS you can have phone calls and
> data connections at the same time.
> 
> And for Bluetooth PAN, this is not exported. Seems that Apple is giving
> out extra status information over mDNS when you are using Tethering, but
> that is not really a standard.

Oh, yet another funky Apple mDNS thing :)

Adding Suspend to the Session API would only make sense if we are
planing to integrate this.

> It would be better to extend the PAN profile with these details, but I
> have not yet seen an updated specification. Except for some APN
> configuration settings and even these were early drafts.

Would it make sense to bring this kind of request to their
standardization group? It sounds like a reasonable thing to do...

cheers,
daniel
_______________________________________________
connman mailing list
[email protected]
http://lists.connman.net/listinfo/connman

Reply via email to