Opinion: The Memory Spot, HP's tiny new data silo, should be a caution to IT systems planners.
Hewlett-Packards new wireless data chip, the so-called Memory Spot that the company disclosed in a press conference in July, squeezes a remarkable number of mistakes into a space about the size of a grain of rice.
Im not talking about the chips ability to perform its intended function. Im sure that HPs legendary research skills have achieved an impressive technical objective, enabling a 10M-bps radio link with up to a 4-megabit data repository (probably more in future versions) that a person could stick on the back of a photograph for about a dollar.
It will be easy to demonstrate this technology in trade shows and on TV commercials, with proud parents swiping a little wand across the back of a snapshot to see a video (from which the snapshot is only one frame) on a nearby screen. Doctors will be shown reading patients histories from a spot on a paper wristband.
As innovations go, this little chipintegrating a capacitor array, modem, loop antenna, microprocessor and memory subsystem into a single componentis a dandy science project.
In microcosm, though, the Memory Spot demonstrates exactly the things that an enterprise IT architecture should have stopped doing at least a decade ago, and that even individual users will soon be moving beyond.
First, the Memory Spot represents the minimal case of the data silo. Every separate device holds its own distinct collection of data, and the wireless link requires the probe to be within about a millimeter of the integral antenna. You wont be able to take an album full of photos and catalog them quickly; you wont be able to take a drawer full of hospital patient files, each with a Memory Spot on its cover page, and swiftly identify patterns of drug interaction or post-operative infection.
You could set up a workflow in which a Memory Spot captures data thats subsequently collected into a larger repositorybut what, exactly, is the contribution being made by the Memory Spot as middleman? Why not just collect the data in a proper database from the get-go via Wi-Fi or another pervasive wireless connection?
Hewlett-Packard is intensifying its R&D efforts. Click here to read more.
For enterprises, my advice is to look at the Memory Spot and go in the opposite direction: Look at your operations to see how many places data is being capturedcreating opportunities for error or loss or inconsistencyand try to shrink that number rather than merely shrinking the hardware.
HP has sought to position the Memory Spots close proximity requirement as a security feature, compared with RFID (radio-frequency identification) tags that can be read, perhaps nefariously, at distances of up to 10 feet. I dont find this persuasive. If Im able to handle a document or other object thats equipped with a Memory Spot, I can embed a probe in a shirt-cuff button to capture data without being noticed.
Its just another variation on the discredited idea of security through obscurity to say that a 1-millimeter limit on access distance is an important contribution to keeping data safe. After all, you need actual physical contact to read the magnetic stripe on a credit card, but there are inexpensive handheld readers that let those rare (but not unknown) miscreant waiters, gas station attendants and other service employees covertly swipe and record stripe data while your card is in their possession.
Enterprises should take this as a reminder that the biggest threat to your data is not the mysterious black-hat attacker, but the authorized party to a transaction who abuses access privileges from inside whatever data fortress you may have built.
You can defend against the black hats with technology, but defense against rogue employees or untrustworthy supply chain partners requires careful thinking about business process design and privilege management.
Finally, HP officials say that the Memory Spot will be able to perform authentication and on-board encryption functions. We saw the same idea in the Java rings (using iButton technology from Dallas Semiconductor) at JavaOne in 1998. Its a good idea to put authentication and encryption close to points of data collection and use, but its a bad idea to use cryptography without a strong infrastructure for managing keys and for assigning (and, crucially, revoking as needed) the privileges that go with them.
Peter Coffee can be reached at email@example.com.
Check out eWEEK.coms for the latest news, views and analysis on servers, switches and networking protocols for the enterprise and small businesses.