Bob Cringely is going against the grain, as usual (and quite refreshingly so, too), in his recent column on why telcos should keep their circuit-switched network. I am not sure how this is done in the US, but at least in Norway, the seemingly circuit-switched POTS network is actually fibre-based and packet-switched in the background, the only circuit-switched bit being the “last mile”, so I am not sure how his arguments would work here. To me it seems, like it does to Techdirt, that this is an argument for telcos hanging on to their dying technology as it is being disrupted – though the “super-customer” forcing the incumbent to hang on to the old technology seems not to be there.
Anyway, the interesting part is Bob’s argument that it should be possible to reduce the bandwidth needed for sending video by mimicking the protocol of the optic nerve, which according to Bob has a capacity of 100 Kbps. Now, I am no expert in optics or anatomy of the eye, but it occurs to me that one of the reasons the eye can do with relatively limited bandwidth is because so much filtering takes place before the picture gets transmitted. When I look at a movie, I don’t take in the whole picture at once – I focus on some part of it, and am only dimly aware of the rest. I do this by positioning my eyes towards what I am looking at and then focusing it – in the process selecting just a few bits of all millions of “bits” the world insists on sending towards me in analog form. What I focus on comes through in glorious detail (at least when I have my glasses on) and the surrounding stuff is out of focus (and there are different physical sensors in the eye to handle this – two protocols, if you like).
Now, since no two people focus on the same part of the video picture through a movie, you either need to send the whole picture with the same quality, or you need to establish some form of two-way communication, so that only what the eye actually will look at will be transmitted towards it. So in other words, to send video down a 64Kbps connection, you need it to be two-way, with almost zero latency. Moreover, you would need one connection for each viewer.
The eye selects what to see, in communication with the brain. The world, which sends images to the eye, has unlimited bandwidth. I may be wrong, but in order to send video over 64K, as Bob proposes, it seems to me we need to extend the selection properties of the eye into the server rather than the send the whole image into each person’s home – and communicating eye-tracking with focus information from each individual viewer back to be processed in time for the image to be selected and transmitted in optimized format seems to me to be a formidable challenge, circuit switching or not.
So, to me it looks like packet switching is the way to go, also for technical reasons. A second argument is that telcos make money when users talk to users, not when they connect to a central feed (that is the domain of broadcasters, and for telcos normally only a way to get users to subscribe to their network, until talking to each other takes over for talking to a central source). And packet switching allows for video-to-video sharing (legal or not). That is the future, and whether the telcos and broadcasters make the transition, I don’t know and I don’t care. I know I, as a user, will. The circuit switched network will not.
Update June 30:
David Isenberg has a rather more direct take on this in his “I[diot], Cringely“.