[asterisk-dev] Re: chan_sip - performance, and OPTIONS poking for 1000s of peers

[Eric "ManxPower" Wieling]

Ryan Mitchell wrote:
> Good point, thanks.  rand() w/o locking is fine, I was just
> copying-and-pasting w/o thinking.
>> Would this code get run for each OPTION packet sent? If so, perhaps a
>> less expensive random function could be found. Or maybe
>> thread_safe_rand is already cheap enough, it just seems overkill for
>> something with such a low requirement for randomness quality.

Whatever is done it would be VERY nice to have qualify smoothing in SIP 
similar to the way it is done in IAX2.