[Asterisk-Dev] We can do better than Skype

Daniel Pocock daniel at readytechnology.co.uk
Tue Oct 19 08:01:43 MST 2004


>>but Skype's protocol is more successful at getting through firewalls and
>>NAT
>>    
>>
>
>How is that? Do you mean to say there are presumably many more Skype
>sessions than IAX sessions, hence they are more successful?
>
>Otherwise, how is IAX not successful in traversing NAT?
>  
>

Skype has chosen to support TCP and sacrifice performance in order to 
guarantee compatibility.  Almost all NAT gateways and firewalls allow 
HTTP (some via a proxy), some allow outgoing TCP, but not very many 
offer good UDP support (or if they do, it is often disabled by some 
network administrator).

>  
>
>>Maybe we need to look at the possibility of IAX over HTTP
>>    
>>
>
>You'd have to write a wrapper protocol because HTTP is text based and
>IAX is binary. This would probably lead to inefficiencies. Efficiency
>is one of the main strengths of IAX.
>  
>

It would be a bit like Kazaa - the protocol would attempt to setup a 
call over UDP first, but if that wasn't possible, it would fall back to 
various NAT and firewall workaround methods, such as HTTP.

The user may also be able to force it to operate in a particular mode, 
or specify a list of 'permitted' modes in order of preference.

>IAX over TCP?
>
>Real-time is not typically a good thing to run over TCP. Call quality
>will probably drop.
>
>  
>
You are completely right - but some users have no choice.  There are 
still many networks (eg hotels) where users have little control over 
infrastructure and they just want to 'plug and play'.  Therefore, if we 
want to give users maximum flexibility, we have to give them a choice of 
operating modes.

If we don't do this, then they will continue to be attracted to 
proprietary vendors who are willing to offer such solutions.




More information about the asterisk-dev mailing list