Network end users to pull down 2 gigabytes a day, continuously?

Responding to postings by Colm MacCarthaigh and Marshall Eubanks:

1. There is practically no live television (at least in the United States).
After the Janet Jackson episode, networks are inserting a 5-second (or perhaps
it is a 10-second, I don't recall) pause, in order to stop anything untoward
from showing up on the screen.

Admittedly, there are live events (videoconferencing, or sports events that
some people get a thrill out of watching in real-time), but that is a small
fraction of total traffic.

2. Business models (such as advertising-financed TV) are certainly slow to
change, as both businesses and consumers do not alter their habits on Internet
time. But neither business models nor consumer habits need to change when
you move from streaming to file downloads. As long as the transmission does
not have to be absolutely real-time (as it does with videoconferencing),
you gain a lot. Say you have a 3 Mbps download link, and the transmission
speed of your video is 1 Mbps, start shooting it down at 3 Mbps (possibly
allowing the customer to start watching it right away), and after 5 seconds
you will have the first 15 seconds in the buffer on the customer's device.
Even if that person has been watching from the beginning, you now have a
10-second grace period when you can tolerate a complete network outage without
disturbing your customer. Just think of how much simpler that makes the

And if you do worry about long videos not being viewed to the end, shoot
them down to the customers in 10-second increments.

This solves concerns about advertising and everything else. And of course
you can encrypt the files, and do whatever else you want.


P.S. I have been puzzled by the fixation on streaming for over a decade.
A couple of years ago I wrote about it in "Telecom dogmas and spectrum allocations,"

At my networking lectures, I often do a poll, asking how many people in the
audience see any advantage (for consumer, or service providers, very vague
requirement) in faster-than-real-time download of video. The response rate
has ranged from 0 to 20%, with the 20% rate at two networking seminars at
Stanford and CMU, full of networking graduate students, professors, VCs,
and the like. There is a (small) fraction of people who see buffering and
file downloads as the obvious thing, and others mostly have never even
imagined such a thing. What's strangest is that the two camps seem to
coexist without ever trying to debate the issue.

Is there really a much of a difference between streaming and buffering?

Is XM radio streaming or buffering?
Is TiVo streaming or buffering? Is AMC Theatres streaming or buffering?
Is RSS feeds streaming or buffering?

Geeks like to debate it, but the difference is mostly one of degrees not a bright line. Usually the data is processed multiple ways. A security camera may be both a linear stream and a random access buffer. A movie
theatre may download a digital movie via satellite file broadcast but play
it as a linear event at pre-determined times.

True believers may like to think there is only one true answer. But it is probably more realistic that the network will need to support multiple different ways of accomplishing the same thing because different people will want to do the same thing different ways.