A couple of days back, Sightspeed announced a partnership with AMD whereby AMD will rebrand Sightspeed client as AMD Live Communicator. This is a piece of good news for them. I used that occasion to talk to their CEO, Peter (who has started to blog) to update myself from the last time. They have introduced a couple of capabilities that I think are worth noting.
The first one is a new feature that they call Sightspeed Web, which allows a Sightspeed subscriber to have Sightspeed session with one who is not. To do this, the Sightspeed subscriber sends a specific URL to the friend, who can download an ActiveX client as part of visiting that URL. The ActiveX client has almost all the features available in the standalone client. This is very similar to what one could do in FWD. This is very useful for Enterprises to add click to (video) call to their web site.
The second capability that is woth noting is their claim that their NAT/FW traversal capability is very effective and can also handle symmetric ones. It is so effective that only a small fraction of calls require relay nodes. Since this is counter to my understanding, I wanted further clarification. The following is the response I received from their CTO, Aron:
“I can’t elaborate much further on how it works, but we basically have the best understanding of firewalls, NAT/Firewall’s and NAT devices in the marketplace. With this knowledge we built a solution which maps the current state of your firewall (symmetric included) when you want to start a call. By knowing the current state we can then build a direct peer2peer connection between the endpoints. The types of firewall’s that require persistent relays are usually ones we have not analyzed in our labs or ones which don’t follow any known patterns. Relays are also used for misconfigured firewalls that limit where destination traffic can be sent too. We find a fair number of Enterprise firewalls incorrectly limit destination traffic by default.”In my opinion, this is a major development that will impact others as well. This suggests that one can further classify symmetric NAT/FWs and solve the traversal problem for a subset of them. Posted by aswath at October 12, 2006 10:30 AM
This is interesting commentary. Sightspeed has not give us any specifics on how they do it. They've just said that they "understand how NATs work" better than anybody else.
If you look at the implicit NAT behavior analysis that was part of the original STUN draft, this statement becomes even more interesting. During the recent update to that standard, the people involved, many of whom are working for Cisco and other huge vendors of networking equipment have basically backed away from that, removing the analytical process almost entirely from the bis update, primarily because they judged there are too many variations in NAT implementations.
While I would very much like to believe that by being smarter about watching the NAT and analyzing its behavior, you can make better decisions about how to travere it, I would remain skeptical however, until there are more details. Certainly if Sightspeed has some IP in this area, they are entitled to it but I'm dubious about hearing the CTO of a startup thats trying to make a big splash in the market saying simply that they are smarter than everyone and then assuming that the problem has been solved. Let them file a patent and then show us what they've done. We'll be able to tell whether they've really made a breakthrough at that point.
Posted by: Frank Miller at October 12, 2006 03:11 PM