[asterisk-dev] [Code Review] SIP: peer matching by callbackextension

[Olle E. Johansson]

21 okt 2009 kl. 22.10 skrev David Vossel:

>
> ----- "Olle E. Johansson" <oej at edvina.net> wrote:
>>>> On 2009-10-09 06:18:23, Nick_Lewis wrote:
>>>>> I cannot think for the life of me why this config option would
>>>>> ever be left off. There will no doubt be loads of people reporting
>>
>>>>> callback extension matching problems who will have to be told to
>>
>>>>> enable matching in the config. Eventually the asterisk team will
>>
>>>>> get sick of having to inform this never ending stream of people of
>>
>>>>> the config option and will flip or remove it. However if having
>>>>> matching as a disabled option enables a consensus to be reached
>>>>> then I am delighted
>>>
>>> Yeah, it's really just about providing another knob of control while
>>
>>> not changing the default behavior that has existed for quote a
>>> while.  I'm good with this approach.
>>>
>>
>> I strongly insist, as before, that we write a doc about all the
>> changes and review them together instead of patching something that's
>>
>> already broken. We now have at least two changes on the table and  
>> more
>>
>> to come.
>>
>> The whole callback extension code is a broken concept that should not
>>
>> have been there in the way it is from start. It's a bad patch without
>>
>> any architecture thinking behind it. I think we'd better fix that
>> instead of adding new code to fix the broken code.
>>
>>
>>
>> /O
>
> I understand you don't like this change.  You've been very vocal in  
> telling us peer matching needs to change and you probably have one  
> of the best understandings of how this should happen.  I think that  
> if you offer some direction in the form of organizing the changes  
> that need to be made, and proposing at least an outline of how we  
> can begin to address this as a whole, it will definitely help us  
> move forward with the discussion.

Have you read the documents I've been refering to many times? The  
"future" architecture for pineapple (www.codename-pineapple.org), the  
current one that I wrote up after the kill-the-user disaster (caused  
by me, myself and propably I).
Those two combined are a good starting point for writing up a new one.  
Easy to find if we had svn view... ;-) Search the list and you'll find  
plenty of pointers. I'm a bit off-line now and can't find them. Maybe  
tomorrow.

Writing this up will take some time, that I currently don't have  
plenty of and no funding for. That's why I keep pushing you guys to  
start doing it with the starting points I've given you earlier...  
Maybe we can set up a discussion on IRC on this topic or schedule a  
conference call.

Regardless, until someone has stepped forward with an architecture  
that we all can approve, I want to put this patch and my patch on hold.

What it's really time to do, is to fork chan_sip and start from  
scratch with this model *without* a requirement for backwards  
compatibility. Keep the old junk and create a new channel that will  
have a realtime system that people understand and that works, a  
configuration set that people understand and that works and  
documentation that people understand... But where on earth will we get  
the time and resources for that?

Not having that time, and not having the required documentation is not  
a good excuse for integrating patches that have a high risk of  
breaking more than they are solving.

Long rant from a jetlagged mind in a cold country called Sweden :-)
/O