[asterisk-users] Some IAX calls do not disconnect.

[John Novack]

Steve Totaro wrote:
> On Tue, Jul 7, 2009 at 7:43 AM, Tim Panton<thp at westhawk.co.uk> wrote:
>   
>> On 7 Jul 2009, at 05:05, Steve Totaro wrote:
>>
>>     
>>> On Mon, Jul 6, 2009 at 9:41 PM, Tim Nelson<tnelson at rockbochs.com> wrote:
>>>       
>>>> ----- "Steve Totaro" <stotaro at asteriskhelpdesk.com> wrote:
>>>>         
>>>>> Just use SIP and solve all your problems.
>>>>>           
>>>> I seem to be noticing a common element to your posts about IAX. :-)
>>>>
>>>> I've been successfully using IAX in a large scale environment with no
>>>> problems... yet. Can you shed some light on the reasoning behind your
>>>> obvious dislike of IAX2? It is supposed to be the 'killer' of SIP from a
>>>> usability standpoint (NAT traversal is quick to my mind...). BUT, is it just
>>>> not robust enough in your experience? Are there inherent problems with the
>>>> protocol itself? Is this changing now that IAX2 has it's own RFC? Is it the
>>>> implementation within Asterisk that is the problem? I'm very interested to
>>>> to know where your disdain comes from. :-)
>>>>
>>>> Thanks Steve!
>>>>
>>>> --Tim
>>>>
>>>>         
>>> First define large scale.  It certainly means different things to
>>> different people.
>>>
>>> Second, It comes from huge amounts of audio problems over many, many
>>> years, and many, many implementations.
>>>
>>> I actually don't have a disdain for it, it has made me a good deal of
>>> money by fixing ITSPs/carrier's audio issues by switching them to SIP
>>> and still does so I have a fondness for it.  Keep up the sub par
>>> protocol, it helps with the balance sheet!
>>>
>>> Third, it will never kill SIP.
>>>
>>> First of all, Digium owns the name and we have seen what they are
>>> willing to do to attack people for trademark or copyright infringement
>>> (think about the Google Adwords debacle and the the Open letter to
>>> Digium drafted by Trixter that I am not sure was ever fully addressed
>>> by Digium.)
>>>
>>> It would have to be renamed or something.  I think the same thing of
>>> DAHDI.  They want control over the the names Inter Asterisk Exchange
>>> and Digium (whatever the heck the rest of it means.)
>>>
>>> Second, SIP is the industry standard.  Only a couple of goofy phones
>>> do IAX2 as far as I know, some crappy handsets I wouldn't even bother
>>> testing if offered as a free demo unit.  SNOM might now, I am not sure
>>> but I think I read interest in it or it was actually accomplished.
>>> SNOM is OK but I was never a big fan.
>>>
>>> When I see it on a Polycom, Cisco, NEC, 3Com, or any other major
>>> vendor's phones or platforms, then I may rethink my ideas.
>>>
>>> If 3Com and Digium are partnered up now, how come the NBX for V3000
>>> doesn't support IAX2?  They do have SIP.
>>>
>>> Second, there are work arounds for just about every downfall of SIP,
>>> like NAT traversal and the like.
>>>
>>> Third, ALL REAL TIME VOICE traffic is on a single port.  There is a
>>> big issue there, I won't elaborate, but just think about it.
>>>
>>> SIP is here to stay until some other protocol comes about, but
>>> certainly not IAX2.  It will be along the evolution of H323 to SIP to
>>> X., but not IAX,lol.
>>>
>>> Do you realize that most providers are dropping IAX2 support, even
>>> IAX.cc recommends SIP, gotta wonder why?
>>>
>>> Maybe it is all good now, but I won't bank my reputation on it.  I use
>>> what I know works well, period.
>>>
>>> Even unnamed Digium Employees have poo pooed IAX2, albeit a year or two
>>> ago.
>>>
>>> It looks good on paper, didn't perform well historically, and now just
>>> like anything that I have lost trust in, it has to earn my trust back
>>> and that is not easy.
>>>
>>> --
>>>       
>> Obviously Steve and I don't agree about this.
>>
>> There are places where IAX can go that SIP just can't.
>>
>> When Steve says just use SIP, what he is actually recommending is
>> to use SIP/STUN/SDP/RTP/IPSEC to get the same result.
>> (at a 50% bandwidth overhead)
>>
>> i.e. replace a single 100 page RFC with something like 100 RFCs :-)
>>
>> In a big organization where you control the network infrastructure, that is
>> an entirely viable solution, but when you want to get calls through a messy
>> network without having to fill out an infinite number of change requests to
>> the firewall team you should consider IAX.
>>
>> The mess that SIP makes is reflected in the number of bugs and the code
>> size.
>> I'm currently working with a SIP stack that is about 10x the size of the
>> comparable IAX
>> codebase, which matters in some environments.
>>
>> As to the 'everything over a single port' issue, this is no longer such a
>> big deal.
>> (And it is exactly this feature which provides IAX's firewall penetration)
>>
>> Most modern Linuxes support multiple threads reading datagrams from a single
>> datagram socket. The current IAX implementation in Asterisk doesn't support
>> it,
>> but that's an implementation issue, not the protocol itself.
>>
>> Also IAX now supports redirecting the media - which could be used to send
>> it to a separate port on the same box.
>>
>>
>> Various Digium employees have also badmouthed SIP (I think we all have
>> after a bad day at the SDP coalface), so you can't take such remarks too
>> seriously.
>>
>> I overheard a senior Cisco employee saying "So you were right all along
>> about IAX "
>> to a very senior Digium employee, which also proves nothing much :-)
>>
>> Competition is a good thing - even amongst protocols.
>>
>> T.
>>
>> Tim Panton - Web/VoIP consultant and implementor
>> www.westhawk.co.uk
>>
>>     
>
> I am not sure what you disagree with.
>
> I can cite several ITSPs/carriers that have had their problems wiped
> out with simply changing Dial(IAX2/,,) to Dial(SIP/,,).  I have not
> asked permission so I will not but it is a reality.
>
> OpenVPN solves a host of issues of NAT and also security for networks
> beyond your control.
>
> I have been forced to use NAT and IAX2 and have had no problems at times.
>
> I have had to penetrate triple NATs and many blocked ports, one on my
> side and then two on the remote side.  The funny part is the remote
> side NAT was done by the provider (East African Monopy).
>
> I am sure they were doing it to prevent VoIP (and other server
> services) which was not legalized until a year or so after we got the
> IAX link up too.  Speex was a lifesaver too but latency was worse than
> a CNN reporter talking to a correspondent in China.  It takes a while
> to get used to not talking on top of people in high latency scenarios.
>  Something about the uneasy silence but with practice, it is nothing
> more than a bit of time waster.
>
> Other times, in real "Large Scale" setups, it didn't perform well.
> Breaking up, static, very poor audio.  These "Large Scale" setups have
> always been in "Controlled" circumstances, even if over the public
> internet.
>
> I never could figure out why or under what variables IAX2 starts to
> sound "bad" but SIP has always fixed the issue providing bandwidth,
> packet loss, and jitter are all sufficient and no other apparent cause
> of audio problems.
>
> Like I said above, in the interest of time and doing what "works",
> find Dial(IAX2/) replace Dial(SIP/)
>
> Like I said, once (really at least seven times) bitten, twice shy, and
> I did add the caveat that my experience was dated.
>
> IAX2 may very well inspire a new industry standard in the evolution of
> widespread adopted protocols, but it, itself, will never become one.
>
> It is sort of like Megaco and MGCP (but not really).
>
> Thanks,
> Steve Totaro
>   

Starting to smell like a religious argument here.
My 2 cents:
A worldwide private network using the IAX2 protocol has proved very 
successful, with the easy setup by less than skilled computer savvy 
folks clearly demonstrates there is a time and place for the protocol. 
Sound quality and packet loss equal to or better than SIP
Lack of devices using the protocol certainly isn't due to the protocol 
itself, but the way it has been (not) marketed .

Choose the proper tool for the job.
Be happy there is more than one .

John Novack

-- 
Dog is my co-pilot