Handling of network namespace IDs

[David Ahern]

On 5/18/17 1:56 AM, Andreas Fett wrote:
> Hi,
> 
> On 18/05/17 03:25, David Ahern wrote:
>> On 5/16/17 3:05 AM, Thomas Haller wrote:
>>> Should the objects become aware of the namespace (and have it part of
>>> their ID)? If not, you would need one cache per namespace.
>>>
>>
>> I think that would be best since all network objects (links, neighbors,
>> addresses, routes) are per namespace.
> 
> Sorry, I don't quite understand. You'd prefer the first or the second
> approach mentioned by Thomas?

The end of that sentence -- one cache per namespace.

> 
> As far as eg. nl_object_identical() is concerned, I think adding another
> attribute there is no problem.
> 
> But without different caches the question remains how for example
> rtnl_link_get() and rtnl_link_get_by_name() should work without changing
> their API?

Yes, exactly.