[asterisk-dev] I finally did it! * IMPORTANT NEWS ON CHAN_SIP IN TRUNK

[Johansson Olle E]

9 jul 2008 kl. 15.31 skrev Leif Madsen:

> Johansson Olle E wrote:
>> After years of preparation the sip_user is gone from chan_sip. We now
>> only have one type of
>> device in-memory data structure in trunk.
>
> Could you elaborate as to why this is being done and how it is
> beneficial? I see that you've mentioned you're doing it, but for the
> benefit of all, I'd like to understand what these changes do for us as
> integrators and users of Asterisk.
>
> The way user/peer/friend is a bit convoluted in chan_sip, but after  
> many
> years of using it, I think we've gotten it pretty stable with some
> reasonable documentation on how to use it.
>
> I'm sure there are good intentions here, but as a non-developer I'm  
> just
> a bit confused as to what advantages we're getting out of this.

For the developer, it means less data structures to maintain and less
memory occupied by double information (the type=friend used to
create both a user and a peer object in memory with mostly the
same data).

It  also means that the call states now have only one structure
to handle, so that we better handle blinking lamps, it's simply
getting easier to handle a state when we have only one
in-memory object.

The user/peer/friend configuration scheme still works as
before. Or should - that needs to be tested. :-)

So far, a lot of simplifications in the source code which will
hopefully lead to a better product. No changes for the user,
more than the possibility to remove the "limitonpeer" setting.
I will add some more documentation soon, but the idea at this
point is to minimize the effects and changes for Asterisk users.

In the future, I want to change the type= field and create better
object structures that work with the SIP protocol. The user/peer
model for SIP is really awful, but that's a separate issue.

/O