[OpenSIPS-Users] [RELEASES] Drafting OpenSIPS 1.10.0 TODO list

[Bogdan-Andrei Iancu]

Hi Saul,

On 02/11/2013 05:38 PM, Saúl Ibarra Corretgé wrote:
>
>>> So, please do not be shy and make your points here ;).
>>>
>>> Also we plan a second round of discussion / selection via an IRC meeting (probably in 2 weeks or so).
>>>
>>
>> You asked for it :-)
>>
>> Here are 3 big items I think we should have in OpenSIPS:
>>
>> 1. Non-blocking TCP operations (this is already on the list)
>> 2. "Outbound" support (RFC5626) - supporting this may affect the design of point 1.
>> 3. WebSocket transport
>>
>> As for bug fixes, there are a few that come to mind:
>>
>> 1. Proper routing of in-dialog messages when GRUU is used. (the dialog should remember the path and loose_route should route the message to the right place even if an AoR is found in the contact, which happens when GRUU is used. This is a problem if a user re-registers while on a call and she chooses a different inbound proxy, as the in-dialog requests would now follow a different path).
>> 2. PUA module should refresh outgoing subscriptions until told otherwise
>> 3. presence_xml should validate incoming PIDF documents (validate them against the schema, that is). Right now it doesn't, and if a client sends a broken document then the aggregated PIDF document will also be broken. Those broken docs should not be stored.
>>
>> I can elaborate on any of the points in case there are any doubts :-)
>>
>
> Oh, since I'm here, let me add one more thing:
>
> - Get rid of the lumps or add a way to apply all pending lumps to a message manually. This would fix a number of issues that can now happen, like using engage_mediaproxy and removing the SDP of a 183 response.
I have on TODO list some fixup to deal with some cases (like adding a 
lump to a section that was removed) - this will solve most of the 
problems . We had a talk on that some time ago ;) - and I just added on 
the list.

To apply changes in realtime will imply a tremendous work (to change in 
all modules), which I'm not so sure it will pay off. A really bad 
compromise will be a function to force the changes to be push, but IMHO 
this is a terrible hack and it will open Pandora's box for script writers.

Regards,
Bogdan