Speaking of Standards

Over on the EComm2008 blog, Lee Dryburgh posted the transcript of a fascinating interview with Jonathan Christensen, general manager of audio and video at Skype. The interview is well-worth a read as Jonathan provides a preview of his upcoming keynote at EComm 2008 with his view of Internet-based communication and talks about advances they have made at Skype with regard to wideband audio and echo cancellation. I do definitely agree with his statement around the improvements they’ve made with echo cancellation on the Mac. Ever since upgrading to the latest Skype, I’ve made many calls with it from my MacBook Pro without any headset whatsoever and have been told the quality has been excellent (and it has been for me when I’m talking to other headset-free Skype users).

Much more relevant to this blog, though, were Jonathan’s statements regarding SIP. At the beginning Jonathan mentions how he originally got very excited by the vision of SIP and ran around stirring up interest at Microsoft where he worked then. But at the end of the interview, Lee asked Jonathan to elaborate on his earlier comments about SIP. This is what Jonathan said:

Yes, so just one clarification - we use SIP. Where, by comparison to the other operators, we are one of the largest SIP users in the world. All of our SkypeOut minutes and SkypeIn minutes traverse the PSTN via SIP interfaces, basically. So, we use it as an interop protocol where we need to.

I think that the vision of the early SIP founders has been largely unrealunrealized [See comments] in the SIP world. SIP is typically just used for these very mundane trunking applications, like the one that we have, or sending calls between two networks and it’s just calls. The vision of multi-modal communications and rich end points has largely failed within the same. I think that a big part of this is that they didn’t pragmatically just solve basic problems like NAT traversal, for example. They also evolved the specification to the point that it no longer had its lightweight appeal. So, we’ll see, SIP will continue to be [the] dominant protocol in terms of this sort of narrowly defined scenarios but I think that, when it comes to rich communications, you are going to see more of this fragmentation. You’re going to see some islands of providers who are just solving the problems. Just making it work for the user and not being religious about the protocol for example.

Has the vision of rich communication over SIP been “largely unrealized”? What do you think? Are his statements true? Or exaggerated?

FYI, if you are attending EComm 2008 you’ll have a chance to hear Jonathan Christensen’s keynote directly. And if you aren’t yet attending EComm 2008, why not?

P.S. For the record, we, too, are huge users of SIP for our connections to/from the PSTN and also throughout our hosted Evolution platform as well as our on-premise Prophecy product. Developers on our hosted platform also get by default SIP *and* Skype dial-in numbers for their applications.

Technorati Tags:
skype, sip, ecomm, ecomm2008, conferences, jonathan christensen

One of the more interesting (to me) working groups within the IETF right now is the “P2PSIP” working group which is aiming to develop ways to let SIP clients communicate on a “peer-to-peer” basis, i.e. without any servers. As stated in the working group’s charter:

The Peer-to-Peer (P2P) Session Initiation Protocol working group (P2PSIP WG) is chartered to develop protocols and mechanisms for the
use of the Session Initiation Protocol (SIP) in settings where the
service of establishing and managing sessions is principally handled
by a collection of intelligent endpoints, rather than centralized
servers as in SIP as currently deployed. A number of cases where such
an architecture is desirable have been documented.

Peer-to-peer is intriguing to me primarily because it does represent a different deployment paradigm than what we are primarily using today for SIP deployments. Today SIP clients register with SIP servers and all the signaling is generally handled by those servers. With P2PSIP, the idea would be that you remove the servers and have all the routing, signaling, etc. handled by the “cloud” of P2P SIP clients. Clients get added and removed to the P2P cloud as they come and go and all the “intelligence” resides in the cloud.

Outside of the world of open standards, this architecture is best seen in voice with Skype. Skype clients connect to each other and route calls and media packets across the Skype cloud. I should note that Skype is not a pure P2P cloud. As was shown by the 2-day outage earlier this year, Skype still does very much rely on servers for authentication.

Will the P2PSIP working group wind up creating something like an open standards version of Skype? Maybe… maybe not… the effort is really only in the beginning stages. (And you can stay up with what is going on at “p2psip.org“.) There are all sorts of security and privacy issues that have to be addressed but it’s intriguing to see. It’s certainly a group I’ll be monitoring and participating in to the extent that I can.

P.S. If you are curious to experiment with open P2P architectures, you can check out OpenDHT.org, an open, publicly accessbile distributed hash table (DHT). Do be warned, though, that this is really for developers…

Technorati Tags:
ietf, sip, skype, p2p, p2psip