[Kde-hardware-devel] Solid::Control not finding local loopback

[David Hubner]

On 22/12/2009 13:06, Sebastian Kügler wrote:
> On Monday 21 December 2009 22:21:45 David Hubner wrote:
>    
>> On 21/12/2009 12:11, Will Stephenson wrote:
>>      
>>> On Friday 18 December 2009 00:55:39 David Hubner wrote:
>>>
>>>        
>>>> Is this because networkmanager does not provide information on the
>>>> loopback device though the dbus interface?
>>>>          
>>> That's correct - AFAIK it was never interesting to me the networkmanager
>>> folks because our main networking priority is getting laptop users on the
>>> internet.
>>>
>>>        
>>>> I can provide a patch for this problem, i am just wondering what the
>>>> situation on the subject is? :)
>>>>          
>>> What's the use case for adding it?  I can see the completeness angle, but
>>> I don't want us to add code without good reason, that would make the
>>> NetworkManager backend (for example) a hybrid of NM's view of the world
>>> and the kernel's, requiring filtering to avoid confusing NM with
>>> synthetic loopback device unique identifiers.
>>>        
>> Its just a case of a consistent API and checking to see if the local
>> loopback is active, to allow the local loopback to be a device of the
>> type Solid::DeviceInterface::NetworkInterface but not providing any
>> information on the "device" in Solid::Control::* could cause problems
>> with people getting the udi from the base libs and using that in
>> Solid::Control. I am not sure if HAL will work without local loopback (
>> other backends might work without local loopback )  but checking to see
>> if the local lookback is active/present would also be helpful. It would
>> probably be possible to check for "lo" in NetworkInterface devices,
>> therefore being valid, but it seems a slight work around.
>>      
> That's not a usecase. :-)
>
> I'm also wary of adding it, since I don't see real value for it, and it might cause
> regressions (think of the lo interface being shown in UIs it previously wasn't).
> Lacking a use-case for it and given the risk of regressions, we can better leave it
> out in my opinion.
>    
Yer, I cannot think of a usecase for it..

To be truthful I brought it up for selfish reasons :) I am getting all 
the networkInterface devices from Solid::Devices::listFromType and 
checking if they are active and if so displaying info. Having lo in 
Solid::Control::* would have made my life easier. Now I have to make a 
design decision, check for "lo" and add static data or get rid of "lo" 
by using Solid::Control::NetworkManager::networkInterfaces();

I hate design decisions, I always choose the wrong one :)

Thanks