[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On Sat, 16 Jan 2010, tom wrote:
Gordon Henderson wrote:I said theoretically - as in everything other than VOIP doesn't have a problem with NAT.On Sat, 16 Jan 2010, tom wrote:Gordon Henderson wrote:On Sat, 16 Jan 2010, Dan Dart wrote:Try having a small one before having a massive one... It will become unmanagable above 4-5 people unless you are very, very disciplined.True - 3 was bad enough but one was in another continent.That's where VoIP has a slight disadvantage over the PSTN - lag/latency.Not if its done 'properly'.Well...The current tendency is to go through a hub or exchange but theoretically once the call has been setup it should be able to fallback to an ip to ip conversation so if you were to talk to someone on the same network 'branch' then the call wouldn't even have to go as far as the exchange. Most stuff seems to be pretty primitive imitations of the phone network - ultimately designed to get some money out of you somehow.You've forgotten 2 things. One is NAT, (and there are ways to get round it - not always successful and less so when both ends are behind NAT which is going to be the case 99% of the time here) the other is that this is a conference bridge - the VoIP *has* to go via the bridge application/device to provide the one-to-many scenario that a conference entails.(There are other issues where the VoIP needs to go via a centralised server such as call recording, transcoding in-audio DTMF signalling, but let's not wory about these here)And even with SIP re-invites (which is what you're refering to), VoIP still has a higher latency than the PSTN. You probably won't notice it when all people are in the same country (or better, on the same ISP), but get someone abroad in the conference and you certinaly will notice it.Gordon
SIP and NAT were designed (RFC'd) about the same time (c1997). At the time, no-one thought NAT would be around for long - we'd moved to CIDR a year or 2 earlier, NAT was just a passing fad until all the router makers upgraded and that was that... How wrong we were! I doubt NAT and ipv4 will go away in the next decade or 2...
SIP is quite curious in that it encodes a lot of endpoint information inside the data portion of the packets - e.g. the IP address and ports... This is used in the reinvite scenario and it's what most NAT routers, even those which think they know about SIP get wrong. SIP is also like FTP in that is uses a command port and data ports - yet another complication for NAT to handle. The worst scenario is 2 endpoints, both behind NAT. Guess what Skype "supernodes" do ...
VOIP needs bringing into the 20th century (sic) so that the 99.9% of simple calls, that dont need metering, recording etc etc.
Maybe. But while people like me can make money out of it, we will...However... To make that work, you need to accept anonymous calls and without some sort of centralised facility (paid for?) to authenticate calls, we're going to get voice spam real soon now... I do not allow anonymous incoming calls into any of my systems - so that rules out eNum...
VOIP is a missed opportunity - its like office software in that it uses the computer to emulate 19th century ideas rather than looking at the problem and presenting a solution based on the technology it has available.And update or complete replacement for VOIP is long,long overdue.
People are only just touching the surface of VoIP, but feel free to invent and deploy something new though...
Gordon -- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/linux_adm/list-faq.html