Thursday, July 01, 2004

 

SIP expert explains problem with SIP

This article in The Register quotes Rohan Mahy, co-chair IETF SIP and SIPPING Working Groups, explaining what is wrong with SIP. Mahy is quoted as saying
There are very few people in the VoIP industry who understand firewall and NAT traversal well . . . Explaining to these people what is technically broken with their proposals and why is an extremely time consuming process, and is repeated every time a handful of new companies start to go into the operational phase.
Exactly! (a) SIP is too durn complicated, and (b) "the VoIP industry" is too narrow a concept.

Voice is just another app on the net. And the "industry" that will bring VoIP to most of us is the Internet Industry.

It is perilously close to Bellheadedness to believe that VoIP needs a special VoIP industry run by High Priests who can't or won't explain its mysteries. In most cases we don't need powerful protocols and special gear to do basic VoIP-based telephony. In most cases, lighter weight protocols (e.g., IAX) will work fine.

[Thanks to Don Jackson for the pointer.]

Sure, there are always places on The Great Network of Networks where light weight protocols won't work. And there are fancy apps like call centers that demand special protocols like SIP. But we open ourselves to disruption if we believe that VoIP is distinct and special, and that VoIP requires High Priests to implement apps.

Comments:
From what I observe, we only have a distinct VOIP industry for the purpose of PSTN interoperability (thus requiring high priests; the PSTN is a weird and scary place for developers.)

That will rapidly devolve to what you're describing - voice is just another app running on a computer or an appliance, looking as little or as much like a PSTN phone appliance as you're comfortable with. Ditto for the naming system to find the address of another user.

Once we transition from PSTN compatibility, VOIP will be ruthlessly Darwinian; those whose voice apps work well with firewalls, etc. will get lots of users. Those that don't, won't.
 
David Beckemeyer here. David, I think you're making a few leaps here. First to blame the innate complexity of NAT traversal on SIP and then to tie the SIP protocol to an elite "VoIP Industry".

Supporting end-to-end peer-to-peer applications over NAT is a hard problem. It is true that there are very few people in the world that are comfortable with NAT technology. It makes even most good networking coders' heads explode. So there is a sort of "NAT elite" out there, but that doesn't have much to do with SIP or a "VoIP elite".

I agree that the "VoIP Industry" concept is silly, but I argue that SIP can help the contrarians. It is complicated. But it is an open protocol, that I can write to. There are hundreds of pages of protocol specifications I can refer to. And as a friend recently said to me "you cannot swing a dead cat without hitting a member of the burgeoning SIP industry." In contrast, where can I find protocol specs for IAX?
 
We already have a lightweight verison of SIP, its called XMPP.
 
To David Beckemayer:

IAX is an open protocol and can download the IAX Protocol Specification by Mark Spencer and Frank W. Miller here ...

http://www.cornfed.com/iax.pdf


You might also find the following link interesting ...

http://www.voip-info.org/wiki-IAX+versus+SIP


Last but no least, I challenge you to actually use IAX for a significant while and then come back to us and declare with a straight face that you are not a convert ;-)

regards
benjamin
 
Post a Comment

This page is powered by Blogger. Isn't yours?