[asterisk-dev] The New CDR system

[zoachien at securax.org]

Russell Bryant wrote:
> zoachien at securax.org wrote:
>> I'm a bit afraid to overload the iax protocol, imho some simple XML 
>> thing might be better ?
>> I feel like the servers already have to process enough information on 
>> the 4569 port.
>
> You must be insane if you think an XML encoded event would be less 
> overhead than a binary encoded event.  :)
>
> Anyway, chan_iax2 would have extremely little processing to do on 
> events.  They would be sent over the network in such a way that they 
> can be directly passed into the event core and mapped to data structures.
Yes, xml would be more overhead, but can the current iax channel handle 
that at the moment ? It's not all that long ago that a register packet 
taking a little long would block all audio packets for a little time. 
Now its multithreaded but im not sure if it's done good enough. (Im not 
a programmer at all, its just based on user feedback).
Don't get me wrong, i hate XML (probably even more than mark does :)

The CDR in iax certainly will have it advantages, guess im just thinking 
too much about fixing the iax channel first. (although i've seen some 
work on that in the SVN lately). Currently a lot of voip providers put 
on their websites or tell on the phone that if you have issues, its 
better to move to SIP as they cant guarantee the quality. (my question 
is how much of this is related to issues in chan_iax and how much of it 
is due to the use of a single port.

Any kernel programmers here to tell us if there is any difference 
between using a bunch of ports like RTP versus 1 port like IAX on a 
kernel level ? why did the RTP people decide to use random ports ? 
(might be a completely different reason).
I guess you all think about getting it to another asterisk (while i am 
thinking to get it into some database with some IAX to DB conversion on 
the DB server)
> Another argument for this generic event passing mechanism is for 
> device state notification.  After I'm finished playing with message 
> waiting indication, I am going to get device and extension state 
> information converted over to this system.  Then, these states will be 
> able to be shared between servers.  One phone will be able to 
> subscribe to the state of a phone that is actually on another server.
This is indeed cool, (we did it once in a bad way, firefly also seems to 
have some special version of IAX to do that). Maybe this should be 
discussed on a higher level, how to cluster asterisk servers. (where 
also info on node availability is passed, server load, available routes 
- including things where a server was down and needs to retrieve all 
info not just take new events.)
>
> It would be really nice for those of you working on IAX2 softphones to 
> be able to support "buddy lists" with state monitoring natively in 
> IAX2, wouldn't it?
>