[asterisk-dev] Reverse Inheritance

[John Todd]

On Nov 11, 2008, at 9:48 AM, Richard Mudgett wrote:

> Tilghman Lesher wrote:
>
>> I was thinking that it should happen exactly once, at the end of a  
>> bridge.
>>
>> So what happens when Dial() creates three channels, and one of them  
>> sets
>> a reverse-inheritance variable, but then is not the channel who  
>> answers
>> and thus never gets bridged?
>>
>> Is this facility going to be *only* applicable to bridged channels?  
>> If
>> so, it will need to be very clearly documented that way. If that is  
>> the
>> case, then the 'multilevel' version could only ever apply to Local
>> channels, since that is the only method of having more than one  
>> bridge
>> in the chain of inheritance.
>
> This idea could be useful for CCBS/CCNR to pass back a call completion
> record ID.  However, that call will never reach the bridged state.
>
> Richard



I too am somewhat uncomfortable have variables "meaningful" only after  
a channel is bridged and terminated - that seems to be very  
inconsistent and will lead to all sorts of strange gymnastics again to  
move variables back up to the parent channel for cases where one wants  
to get preliminary data before the end of a bridge, or even before the  
bridge is set up.  If the call is "split" into multiple different  
channels (i.e.:  Dial(Local/foo&Local/bar) ) then I think I would be  
comfortable with values changing immediately upon their being set in  
the child channels, and the "last in" will be the current value.  If  
two channels "fight" over a value and re-set it back and forth, then  
that is OK - perhaps that is desired in many cases, actually.

Give the dialplan administrator some rope here - they can make a  
hammock as well as a noose out of it.  If there is truly a desire to  
set different values in different children, then use an identifier of  
some type and let each channel set their own named value based on $ 
{EXTEN} as part of the variable name - pretty trivial to do.  In any  
case, I think the corner case is split channels, but the majority of  
the time it'll be to single channels, and having values set only at  
the end of a bridge will I think be frustrating to the usefulness of  
this method in many cases.

PS: The "+" and "-" leading character idea seems to have few  
proponents; I really don't care what way it's indicated.  "^" is fine  
with me.

JT

---
John Todd
jtodd at digium.com        +1-256-428-6083
Asterisk Open Source Community Director