[Asterisk-Dev] IAX2 trunk really should send timestamps as part of iax2_meta_trunk_entry..

[Steve Kann]

On Jan 20, 2005, at 8:44 PM, Matt Riddell wrote:

> Andrew Kohlsmith wrote:
>> On January 20, 2005 08:04 pm, Matt Riddell wrote:
>>> Wouldn't this reduce the badwidth savings offered by a trunked 
>>> connection?
>> Yes.  So you have trunkcompat=yes and it doesn't send the meta event 
>> with the per-channel timestamps, and your trunking is exactly as it 
>> is today.
>

I'd probably call the option "trunktimestamps", and make the default be 
yes.  It would, in theory be "compatible" either way (the timestamps 
would just be ignored by older implementations).

> Cool. Makes sense.
>
> Would you have the timestamps as difference from main one?

I think I'd store the timestamps as they would be in the regular, 
non-trunked case, where they are the low 16 bits of the full timestamp. 
    I was thinking of storing them using just 8 bits, as the "difference 
from the trunk timestamp/10", so, you'd lose a bit of precision, but 
save a few bits, but then if the timestamps skew, you end up with all 
kinds of wraparound issues that make interesting bugs.  (also, this 
would probably break lpc10 with it's 22.5ms frames :)).    For 400bits 
per second per call, I don't think it's worth the headache and extra 
code that would be needed..


> That way the amount of data would be even less...assuming the 
> timestamps are kinda close.

If you want to save even more, use a slightly larger trunkfreq; like 
60ms or so: Right now,   this might or might not work depending on what 
you're doing, but with this change, it should work fine.

-SteveK