clear: Very shortly, we will all have the option, probably
within arm’s reach, of talking on the phone while seeing
each other.

Think about it: More and more PC monitors and laptops
come with a built-in webcam. More and more smart phones
come with cameras on the front.

Chances are high that even your mother has heard about,
and probably used, Skype.

Then, last week, Cisco Systems introduced a residential
version of its “telepresence” line (see story at bottom left),
called “Umi” (pronounced “you me”). If you’ve seen the enterprise
version, you know that it’s freakishly high in video
quality. (As in, forget about multitasking during that long
video conference call.)

Putting aside the matter of whether we really want to see
each other every time we talk on the phone, there’s another
consideration: upstream bandwidth.

A brief refresher: Cable bandwidth is segmented into
two chunks: upstream and downstream. So far, the big
work in bandwidth optimization is far more focused on the
downstream path (toward homes) than the upstream (from

This makes sense, especially when you consider how
many different ways operators need to carry the same

ESPN, for instance, might be carried in analog, standard-
definition digital and in high definition. It’s entirely
likely that mainstream networks will be carried linearly in
IP, too. That’s each big network, potentially being carried
four ways.

The good news is, optimizing downstream bandwidth is
comfortably provisioned with options. These days, operators
needing more shelf space (name me one that isn’t) respond
by harvesting analog channels, recycling bandwidth
using digital video switches, or building more bandwidth
on the top (1 Gigahertz).

No such panoply of options exists for the upstream
path, which represents about 5% of total available bandwidth.
It’s completely hemmed in between 5 Megahertz
and 42 MHz.

This upstream squeeze issue isn’t new. Twenty years ago,
a lot of the engine room was about building a second upstream
path, way up high, around 1 GHz.

Today, the options to optimize the upstream are still pretty
much the same as they were in the early ‘90s: Figure out a better
modulation scheme that can survive down there (it’s noisy
in the 5-42), move the spectral boundary above 42 MHz, or
build a second path.

And with all these gadgets coming that let us see each
other from far away, it’s probably time to dust off the options
and try again.