[Nice] Gracefull fallback, renegotiations?

[Kai.Vehmanen at nokia.com]

Hi,

btw, sorry that I have not had time to test your branch
yet. A test with sofsip-cli/nice_tester on a live network
would probably be a very useful exercise.

On 03 July 2008, Olivier Crête wrote:
[new nice_agent_fallback() method]
>I have an implementation of this API [1]. I see it replacing 
>the current
>set_selected_pair() (at least in farsight), is there any other 
>reason to keep the full set_selected_pair ?
[...]
>[1] 
>http://git.collabora.co.uk/?p=user/tester/nice.git;a=shortlog;h=refs/heads/force-candidate

In theory you might want to pick a different pair than what
was chosen as the best pair by standard ICE preference 
algorithms. But a better way to influence pair selection
is to adjust the local candidate preference values. 

As having two alternative mechanisms in the API, to do 
essentially the same thing, is confusing (and pain for
library testing), I'd vote for removing set_select_pair()
once nice_agent_fallback() becomes available.

br,
-- 
first.surname at nokia.com (Kai Vehmanen)