[Asterisk-Users] Re: How far is IAX to be a Standard

[Karl Brose]

Benjamin on Asterisk Mailing Lists wrote:

>IAX is so vastly superior to SIP, that the comparison shouldn't be
>things like VHS versus Betamax, but it should be more like horse
>carriages versus motorcars.
>

And what to you base such an assertion on?
Would you care to elaborate on the technical justifications?
It appears you are in the habit of making similar sweeping and
unqualified statements regularly.
In time you should learn to be more objective.

The reality about IAX vs SIP looks much different.
SIP quickly replaced H323 in the market of voip endpoint devices
and call termination/origination because of its design simplicity,
transparency and easy debugging potential because it's text-based and
built on existing HTTP protocols. H323 is a complex binary protocol,
albeit very powerful, but difficult to implement. Today it is still
used in carrier interconnects and transmissions, where it is dominant.

SIP, in its foundation, is a state-less protocol alleviating network
elements from maintaining state information. It is very light-weight
and facilitates call control independent of media transmissions.
It's unmatched so far in flexibility and expandability.  That is a
strength.  Because of its simplicity, it has been quickly adopted
by device manufacturers who found it easy to expand and adapt to a
changing market place. The latter has added a host of new features
and added quite a bit of complexity to todays SIP implementations
and protocol.
The rapid growth of consumer market of high-speed Internet access
and the simultaneous lag of (complacency in) implementation of new
addressing protocols (IPv6) created the unfortunate proliferation
of NAT devices, which truly is the worst thing that happened to voip.
The reality today is that SIP has its problems in this environment,
but so does H323.
IAX and Skype are different. They are peer-to-peer, state-full protocols
that integrate control and media transmission in the same data stream
and communicate on a single, well-known UDP port. They are therefore
more easily routable through NATs. This is not a negligible advantage,
and can be a powerful technology enabler at the present time.
Other advantages of IAX include increased efficiency in trunk
(multiple data stream/call) transmissions, because of its compact
design, and perhaps some additional peer-to-peer information exchange.
The trunking bandwidth efficiency is only a marginal advantage in
situation where consumer NAT devices are encountered, since most such
installations don't require the bandwidth on one hand, and the current
limitations of consumer cable/DSL bandwidth are no doubt temporary.
If bandwidth were such a concern, we wouldn't be running TCP/IP anymore,
but had switched to ISO protocols 10 years ago.
The binary nature of IAX protocol, while no doubt yielding efficiency,
is not a recipe for rapid expansion, it will be harder to match SIP's
open architecture and ease of debugging.
IAX appears to be a nice complementary protocol in the overall voip
world.  Because of its integration of call control and media it still
has to prove its scalability to large voip systems.  Because of its
current tight association with Asterisk and its capability to exchange
dial plan information between Asterisk systems, it has a clear advantage
in the interconnection of Asterisk servers.

IAX's potential for being a standard is still to be seen, a first step is
a clear documentation and publication of specification that vendors
can rely upon to achieve interoperability.  In the end, however,
standards are not declared by publication or committee votes, but by
the numbers of implementations.  The activity surrounding IAX and
Asterisk at present is encouraging and stimulating, no doubt.
But there is other competition as well, BT is implementing not SIP,
but MGCP in their transition to an all voip-based network.  Such potential
momentum in deployment cannot be ignored.

Currently, I don't see how IAX can achieve the kind of flexibility
and versatility of SIP, but let the market decide.  I like using IAX
for certain tasks and will continue to explore its uses.

Your zealous attitude does little to promote qualified technological
exchange on this list or elsewhere, it rather reminds us of ideological
battles between close-minded parties engaged in power struggles.






>
>Besides, only five years ago, SIP was the underdog and H323 was all
>the rage. At that point people were very sceptical if anybody would
>even implement SIP because it was yet another standard created by a
>committee, many of which had failed to catch on in the past. In this
>respect, IAX has a clear advantage in that it isn't designed by a
>committee and that it has already got a significant installed base.
>
>
>>At VON, only a few people even understood what Asterisk was, let
>>alone even had heard of IAX.
>>
>
>Well, you are of course entitled to your opinion, but some folks with
>far more clout and credibility that you and I and pretty much
>everybody on this list seem to think differently. A few weeks ago
>there was a statement from an open source guru at an event making the
>mainstream news all over the world and the statement was this: "Watch
>out for Asterisk, it will be bigger than Linux".
>
>This wasn't coming from an Asterisk zealot driven by wishful thinking.
>It was a guy with enough clout to make top headlines in the mainstream
>media.
>
>Of course IAX, once going RFC, will have to start its own life outside
>of Asterisk.
>
>
>>My only point is that we can't just rely on a better design to somehow
>>magically win out. Getting Digium to create a standard with input from other
>>vendors would be a huge plus and help pave the path forward.
>>
>
>I totally disagree. IAX is not a standards committee protocol. It's a
>grassroots thing. If it catches on, and we have any reason to believe
>that it will, given how far it has already come, then it will be
>through grassroots implementations. It will be through small vendors
>who seek an edge and use IAX to get that edge, forcing other small
>vendors to follow. Once there is a critical mass, even the bigger
>vendors won't be able to ignore it.
>
>rgds
>benjk
>
>