p.s. just pasting a couple lines of text from the article, for those who are wondering what this is after looking at the comments, my typical "use case".
In general, it is illegal to broadcast random signals on
spectrum reserved for TV (or cellular) channels. However, battery-free backscattering devices (e.g. RFID tags) are unregulated and not tested by FCC because the emission levels from such devices is very low  and because they are only modulating their reflection of a pre-existing signal rather than actively emitting a signal in re-served spectrum. Ambient backscatter also falls into this category, and would therefore be legal under current policies."
And the tolerance levels are not set at _zero_, but at a specific decibel level on a specific measuring setup. (I've been researching this professionally lately, especially ETSI EN 389-401). Pretty much all high-speed digital electronics in plastic cases will be radiating on the TV band, but at an extremely low level that passes the regulations.
This system modulates by switching its antenna between grounded and ungrounded. At no point does it inject transmission power into the antenna. Your house is already full of small metal objects that are slightly resonant in the TV band - such as cutlery. You could probably replicate this experiment by attaching a switchable ground to a fork and holding it _right next_ to a TV antenna. Thanks to the inverse square law the effect is so weak that it really does have to be right next to the antenna; in the paper they report no detectable effects at a distance of more than a few inches.
* Not only the FCC, but the NAB [National Association of Broadcasters] will be on this pronto.
* So, consider either:
___o Allowing the FCC to retain a channel for a few years of ambient backscatter experimentation and evaluation (funding a set of transmitters for a while, experimenting with what transmitter signal structure might allow wider ambient backscatter bandwidth and node-to-node distances), then offer it as a tax-supported service ( as with GPS ), or sell it, else
___o Having the FCC allow an interested private owner of TV channels privately experiment with ambient backscatter engineering as proposed (locally, multipoint-piggybacked over 8VSB TV), develop the technology, then license it.
* In no way am I intending to cast aspersions on the authors' achievements here; instead I am very, very impressed! It shows indefinitely durable (battery-less) RF networking having practical applications drawing only half a microwatt for 'transmission,' carrying a general purpose programmable microcontroller on board, using a passive method to modulate an ambient RF signal with a man-made signal below the by the intended (TV) signal. And using analog components where they give significant power advantages, while retaining programmability sufficient to support a networking stack. Wow.
Disclosure: I work in support of the Broadcast TV Industry
Edits: (1) Tax-supported is not 'free' (2) Clarify that ambient backscatter is multipoint, 12 orders-of-magnitude lower power, and distributed, in contrast to the single point of origin of the 8VSB TV transmission
The regulatory question is certainly an interesting one. Is it 'interference' to reflect very modest amounts of an existing signal? Multipath is already a very real phenomenon; if these devices do not appreciably affect transmission, then are they truly interfering?
Put another way, are the regulations written from the point of view of effects (degraded transmission for the spectrum licensee), or in absolute terms (no communication at a given frequency?)
>Reading the paper, what do you think is 'massive overhype'? They demonstrate working devices on the 1-2 foot scale.
To be a more effective network device the one thing you will have to do is (at least) try to increase this signal efficiency. That spells future regulatory problems.
>Multipath is already a very real phenomenon.
Again, interference management is already handled within devices. But now you want to add more interference into the environment? See what I mean by: good luck with that in a court.
>if these devices do not appreciably affect transmission, then are they truly interfering?
Yes, they are still interfering. And the problem only grows as you scale this "new" network.
And to answer the final question: Again, you would have to see what lawyers will try to argue.
"Legality: In general, it is illegal to broadcast random
signals on spectrum reserved for TV (or cellular) channels. However, battery free backscattering devices (e.g. RFID tags) are unregulated and not tested by FCC because the emission levels from such devices  ... Ambient backscatter also falls into this category, and would therefore be legal under current policies.
 New policies for part 15 devices, FCC, TCBC workshop, 2005."
A little relevant data from the paper: They do a somewhat crude experiment in the paper looking for corruption of TV signal, and find no effects except when the TV antenna is within a few inches of the device.
Which of course shows that these devices do indeed interfere with the normal reception of the signals they are parasitizing.
EDIT: I am guessing not. I imagine they used real-world backscatter to test this out.
This project is trying to figure out what the equivalent of an rfid tag looks like with that power model. They come up with a scheme of being able to absorb versus reflect the ambient signal to communicate between unpowered tags. Maybe the most productive way to think about this is as a step into the larger research area of how to effectively harvest energy and do useful stuff with devices that don't need batteries.
The method was to build a corner reflector out of three orthogonal conducting sheets, but to connect one of the sheets to ground via a PIN diode switch. By turning the diode on and off, the reflection coefficient of the corner reflector could be modulated, in turn modulating the reflected signal.
I'm not sure what the outcome was, though it must have gotten to a certain point of success, since I've memories of a car driving around with a corner reflector on it.
The same student was experimenting with using a plasma as the modulated reflector, but that one ended when he connected the wrong end of the vacuum pump to his gear, accidentally pressurised it, and blew it up!
Not meaning to be critical, rather just want to clarify that our modern-day gigawatt-scale energy grid yields an enormous 15 orders of magnitude more electrical power than the microwatt-scale "harvesting" of TV signals can. That's an enormous power spectrum, shown to have a newly useful application (batteryless mesh networking) at microwatt scale.
http://en.wikipedia.org/wiki/Crystal_radio (no battery needed for wireless reception)
Using that concept to also produce transmissions is the kicker here.
Now, "the walls have ears" takes on an entirely new meaning in "The Internet of Things"...
What they demonstrate, however, is transmission from the 'receiver' to additional devices by modulating how much power is absorbed or reflected.
This is fundamentally different from just using the harvested power to run a second transmitter, because modulating backscatter is 'orders of magnitude' more efficient.
If you could just stick a rice-sized sensor wherever you want and not worry about battery life we'd all have 'smart houses' by now.
Presumably range can be increased, specially if the principle is adapted to use that WLAN's 2.4ghz stronger signal instead of VHF.
They are not harvesting power from ambient RF and using it to power a second transmitter. Instead, they are using that harvested power to communicate with other nodes by modulating their device's reflection of ambient RF, a process that they claim is ~100x more power efficient.
Not an RF engineer, can't judge novelty, but this seems really, really clever.
EDIT: added text in italics above to make my post clearer.
EDIT: I should say that yes the technique is using analog conditioning circuitry to get the job done without storing power in a battery.
This appears to be the major innovation in the paper. As lots of commenters have pointed out, harvesting power from ambient RF is nothing new (whether or not a battery is involved).
(Section 3.2): "By modulating the electrical impedance at the port of the antenna one can modulate the amount of incident RF energy that is scattered, hence enabling information to be transmitted."
(Section 8a): "backscatter communication is two orders of magnitude more power-efﬁcient than state-of-the-art radio communication")
Is the significant value in this it's ability to function at low power?
On a distributed level, there's a lot of other cool stuff that could be done as well.