On Mon, 2009-06-08 at 16:16 -0500, Jud Craft wrote: > Besides the fact that Skype is closed source, I was under the > impression that the idea of distributed voice transmission was an > advantage. > > Shouldn't a ideal communication system have no single point of failure > (or a country club of them, ala a group of ISP servers?) I am curious. What are the points of failure when using SIP or Skype? For SIP, if I wish to have a voice call between two parties, each party must be on the Internet and reachable by the other party. I would assume each party must also be able to reach the SIP server. If there is a breakdown in the path between the two people, or between a person and the SIP server, won't the Internet attempt to route IP traffic through other paths? If either party or the SIP server can't reach the Internet, there would be a breakdown. I have not looked closely at the SIP protocol. There are things I do not understand. What exactly does the SIP server do? I thought the SIP server acted as a telephone directory where one could register one's presence and look-up the presence of others. I thought the SIP server was used to do call setup and call tear-down between the two parties through a control channel. I thought this control channel could use either TCP or UDP. I thought the actual voice traffic went over a separate data channel, and, if we ignore NAT and conferencing, could be between the two parties. I thought the control channel, with the SIP server, needs to be maintained while the data channel exists. I may never know much about the protocol Skype uses because it is a proprietary protocol. I am guessing Skype must have a server where a party registers. I am guessing the Skype server must do call setup and call tear-down. I am guessing the actual voice data is on a separate, data channel. What exactly do these super-nodes, when using Skype, do? Do these super-nodes only route the voice data -or- do these super-nodes also do directory registration and directory look-up and call setup and call tear-down? Can someone more knowledgeable tell us what is correct? On a personal note, I couldn't get friends and people in a company I used to work for, to switch from Skype to SIP. I tried. They said, "Skype just worked". I had to use Skype, on a Windows PC, provided by the people I used to work for. On my personal PC, which runs Linux, I don't have Skype installed. I tried Ekiga on my personal PC, but had problems. My registration with the SIP server would time-out after one hour. I tried changing UDP session time-out values, in iptables, with no success. I suspect, but am not certain, my iptables firewall rules are acting as a NAT/Firewall. I believe iptables keeps a UDP assocation between me and the SIP server when I register. I believe iptables is timing out my UDP association with the SIP servers which, in turn, is causing my registration, with the SIP server, to fail after one hour. I tried Twinkle on my personal PC, with greater success. Twinkle sends periodic keep-alive packets. I am guessing these keep-alive packets keep my iptables UDP association alive. Whatever Twinkle is doing, my registration with the SIP server does not timeout after one hour. I don't run Twinkle often on my personal PC. It does no good running a VoIP program if you have nobody to talk with when you run that program. I have a final question. My cable company is pushing VoIP. Other companies, like Vonage, are pushing VoIP. I thought their VoIP was SIP. Doesn't that mean SIP can be made to work? Sorry for my long-winded messages. I need to learn brevity.
Description: This is a digitally signed message part