Could it be true?
Is it like a movie where at the end everything that seemed to be true was false and forwards was really backwards?
Maybe…
Just maybe…
VoIP is a mistake.
An attempt to make penguins fly, or cats swim.
It can be done. But it’s not pretty or natural or indeed very successful.
Perhaps what we really needed was a “dumb circuit” network. One that can set up point-to-multipoint time-division circuits, potentially with different levels of capacity. But one that has no concept of calls or call routing. All that intelligence lies with the “edge”, and the edge devices use IP to perform any signalling they like.
In other words, we threw out the media plane baby with the control plane bathwater.
Oh dear.
What a horrible thought.
At least we can still have VON conferences. “Voice off the Net”.
We’ll be doing it anyway, since “Video off the Net” (and onto content deliver networks, physical media and broadcast) is already underway too.
Parting thought: if you burn a heretic pundit, should you pay for carbon offset? Or is the hot air saving alone a net environmental benefit? :)
Posted by Martin Geddes at 02:20 PMTrackBack URL for this entry:
http://www.telepocalypse.net/cgi-sys/cgiwrap/mgeddes/MT/mt-tb.cgi/868.
Hilarious.
I have often joked that some day in the future, we, the telecom industry, will realize that the care and feeding of a voice session over the internet is totally out of whack compared with the bandwidth it takes. As ‘regular’ data traffic takes over, supporting real time voice becomes more of a pain in the rear for those who guide the internet. Eventually voice will be banished from the internet for being too much of a hassle.
We’ll envision a new network, one with direct, point to point connections and very low latency. A new network which does not have individual packets to be lost or delayed. A new network designed with no thought for packet data, and thus being optimized for voice.
By then we’ll wonder where all those class 5 switches went...
VoIP, or at least its media part in its current form is essentially an attempt to take TDM voice as is and put it on top of best effort service. Not going to work well. Another problem is that we have quite limited choice of flow control for IP:
- no flow control at all in UDP - bad, no feedback from network
- TCP flow control with a number of problems - can't estimate path capacity quickly, doesn't distinguish different causes of packet loss, have tendency to fill up queue on bottleneck link and as result introduces large delays and non-zero loss even on ideal links.
What might work is voice/video application(s) designed for best effort packet transport from ground up - using highly variable bit rate coding to adapt to available bandwidth and erasure coding to compensate packet loss + flow control that can detect change of path characteristics in time an order of RTT and inform application. Building blocks already exist or in research phase - AMR have highly variable bit rate (not sure if can adapt in ~200ms though), XCP and PCP are trying to solve TCP flow control problems.