[Kamailio-Devel] Moving forward to Kamailio 2.0: Async core design

[Adrian Georgescu]

Hello,

With the risk of stating the obvious, the future success of this  
project is dependent on its ability to reliably deliver larger  
payloads than a basic voice call setup. Presence is one example. IETF  
moves towards removing UDP as allowed transport in the next version of  
the SIP protocol. UDP fragmentation at NAT-ed routers is the most  
visible culprit and there are other reasons.

The way to move forward is to switch to using a reliable transport  
like TCP. This is not reliable with the current design, which is  
blocking at every step waiting for an answer from DNS, Radius, SQL  
queries, and TCP putting it into its knees at every connection  
timeout. We are looking at rebuilding the core to be an asynchronous  
reactor that does not block during the processing of SIP messages.

This means a new design and a lot of work to implement it. Here is an  
open question at this stage to get the big picture so that at least we  
know where we stand as a community on this topic:

1. Do we see this target as high priority for a next generation  
Kamailio 2.0?
2. If yes, what timeframe do we see realistic to achieve it?
3. Are you as a software developer or company willing to put any  
energy into it?

Other thoughts on this matter are welcome.

Thanks,
Adrian