Just back from my weekend trip to Harvard’s Berkman Center for Hyper-Public — nicely-framed précis of my talk here — and already neckdeep in the eternal stream of work. Here’s what’s going on in the studio:

- The update I promised on PERRY:

I sure hope they’re more responsive when it’s Jeff Bezos on the other end of the line, but display vendor E-Ink turned out to be fairly recalcitrant in communicating with us. They just didn’t seem to have any interest in helping us work with them, and when it finally did turn up, the $1500 dev kit we ordered from them turned out to be a pretty paltry affair. (Insulting, actually: literally held together with tape.) Figuring all of this out cost us the better part of a month and, unless we can return the kit, that fifteen hundred bucks.

For the purposes of prototyping, then, we’ve decided to turn to an alternate supplier, Taiwan’s Winstar, who have happily turned out to be much easier to work with. We got their display, we got their documentation, and that let Benedetta and Justin move on to the next, “Hello World” stage of powering the e-paper and getting it to render arbitrary text. (You can see an image of that here. As I say, it doesn’t prove a damn thing developmentally, as this is basically a result you’d expect at the end of Day 1 in a p-comp course, but it makes me happy.)

The display doesn’t operate on its own; it needs to be driven by a microcontroller. We now face the task of sourcing (or, eesh, developing) a microcontroller that does what we need it to do vis à vis the display, no more and no less. Most of the off-the-shelf ones available are wild overkill for this application, and have dimensions and power-consumption profiles to match.

And power consumption, as we’d been expecting, turns out to be the critical variable. Remember that this device needs to be powered up by a ~300 ms pass through an RFID induction field. There’s still some daylight between the amount of power we can pull off of a commercial-spec RFID reader in that interval — say, the one found on a Metlink bus — and the minimum we need to drive the microcontroller and display. The differential isn’t as stark as some folks we’ve talked to had warned us it would be, but it’s not zero either, and that’s what we need it to be.

There does remain the option of including some kind of thin-film capacitor or even battery in the device, which would resolve the power problems — but only at the penalty of a more extensive bill of materials and therefore a higher per-unit cost. My concern is that going that route would push the eventual Farevalue offering over the threshold of economic feasibility, even at scale. (It’s the “Blue Monday” problem, really: such a version might offer a compelling user-facing proposition, but be a money-loser with every unit shipped.) It’s also, frankly, just a less elegant solution. The revolution will be technically sweet, or it will not be at all.

So there’s undoubtedly some real engineering left to be accomplished. Happily, though, we don’t have to solve all of these issues just yet, let alone the challenge of ensuring they all fit nicely in a volume not much deeper than the .76 mm provided for under ISO/IEC 7810‘s ID-1 standard. That day will come soon enough, but for now I’m simply well pleased that we’ve successfully jumped through the first (and widest) of the several flaming hoops we face.

— I wanted to give you a quick round-up of stuff from Hyper-Public, but due to space and time limitations I’ll confine myself to noting that by far the most useful talk, from my perspective, was that delivered by Latanya Sweeney, director of Carnegie Mellon’s Data Privacy Lab. I wish she’d had a longer slot, honestly, because the issues she raised were so apposite, so timely and so immediately applicable to our work that they really could have used an airing in greater detail.

Sweeney’s main message was that, in the age of Big Data we now live in, certain important pillars of data-privacy praxis — de-identification and informed consent foremost among them — have come under profound and sustained attack. (De-identification is the doctrine that data gathered from human subjects can in principle be anonymized, or scrubbed of personally identifying indicators, while informed consent is the standard in law that requires those seeking acquiescence to certain classes of procedure to explain the risks and benefits in language the subject can understand, and receive explicit permission to go ahead.)

I’ve been insisting, at least since Everyware (2006), that when you have tools like online relational analytics engines readily available, there can be no such thing as an “anonymous dataset.” This point was underscored just a few months after the book came out, when reporters working for the New York Times were able to reconstruct identities from ostensibly anonymized AOL search queries, and do so with relative ease. I’m delighted, then, to have Sweeney making this argument far more rigorously than I would have been able to, and from a bully pulpit more prominent in every way.

She went on to point out, though, that informed consent is equally untenable. Anybody who’s ever scrolled immediately to the bottom of a 45-page EULA to click “Accept” understands the weaknesses of the paradigm. And of course it only gets less practicable in anything like the spyscreen scenario, where the collection attempt is made at a distance of meters, concerns several subjects at once, and takes place in an interval that is, again, best measured in the hundreds of milliseconds.

The impossibility of meaningfully informed consent in this context is one of the reasons why we believe spyscreens should be heavily regulated. Its difficulty even in other, less highly-charged circumstances, though, is at least as troubling, because there’s no clearly emergent alternative paradigm that might provide users with some protection while still facilitating data-dependent functionality. This is absolutely not an abstraction to us — certain Urbanflow use cases surely involve collecting data from people using the service — and we’re going to have to find some workable modus vivendi if we’re going to move forward with that initiative.

— You’ll probably have heard tell by now if you’re a networked-urbanist type, but Twitter chose to reassign the @towerbridge account I’ve been talking about for the past few years from original developer Tom Armitage to a marketing agency. Read more here about why I think this is a very bad precedent.

— Logistricks: M is onsite with Nordkapp, Jeff Kirsch starts tomorrow (!), and we’re excited to announce that ace developer J.D. Hollis will be joining us over the summer to work on Project LAFAYETTE. I’m on a plane for Paris this Friday, so if there’s gonna be a FRIDAYS AT 7 it’ll have to happen through your own good offices We’re on again! Details here. See you soon!