Thursday, February 10, 2005

fyi Databases Can't Handle RFID


Pointer to article:,10801,99589,00.html?source=NLT_MW&nid=99589

Kobielus kommentary:
Good ol’ John Fontana of Network World called me the other day for analyst insight on this big bucket of abstraction called the “identity of things.” John’s one of the best tech trade press reporters. You can tell he’s good. He knows the right analysts to call. ;-)

Anyway, I’ve been meaning to opine blog-wise on the “identity of things” for some time. I just didn’t have a good enough fyi handle to hang my thoughts on till I ran across this story this morning.

First off, what the heck does the “identity of things” refer to? On one level, it sounds like some metaphysical plane of existence, some mythical spirit world, some platonic ideal, like the “secret life of plants” or the “lifestyles of the rich and famous.” Like animism: the identities/souls of the inanimate starstuff from which we’re all, magically, composed.

But I digress.

John called for assistance in helping him narrow down the scope of his contemplated article on the “identity of things.” He already had a good understanding of the broad scope of the term, in terms of concrete, real-world, commercial technical approaches, such as IP addressing, RFID, and ID dataweb. He had already ruled out discussion of IP addressing (nothing terribly new there—who’s up on IPv6?) and RFID (far too much into the supply chain management angle than John cared to delve in that particular article).

If I had read this article before John called, I would have suggested he explore RFID’s demands on the data storage, transmission, and processing infrastructure. Yeesh—listen to what one British RFID consultancy is finding in the manufacturing industry:

“[The consultants] contend that the [parts-tracking data management burden] will only get worse with RFID, which will balloon the amount of data that's generated and make indexing the information in a relational database prohibitively expensive and all but impossible. [One of them] estimates that if Wal-Mart Stores Inc. logged all of its inventory via RFID tags for a single day, it would reach 7 million terabytes of data. Gulp! Can your database swallow that? Didn't think so.”

That’s one of the big problems with the “identity of things.” There are just too many “things” in the universe. Try giving every star in the sky its own unique name, including the billions upon billions embedded in galaxies, and don’t forget to give each of the countless galaxies their own unique names. After identifying every discrete point of light uniquely, now try storing and managing all those names (plus the associated descriptive attributes of each star) in some master directory database in the sky. Clearly, the directory itself would have sufficiently massive gravitation to form its own black hole, sucking all of the named “objects” in the universe down into some freaky meta-universe, never to be heard from again.

But I’m rat-holing on a metaphor. Also, I’ve overlooked all the planets, asteroids, comets, and empire deathstars orbiting all those celestial bodies. We would need a universal Unicode that supports Klingon character sets, at the very least, to do justice to all that heterogeneity.

Actually, John was primarily interested in ID dataweb—aka federated resource sharing environments built on emerging Web services standards, especially Extensible Resource Identifier (XRI) and XRI Data Interchange (XDI). That’s a hot and interesting topic. John just wanted to know—as do we all—when it will actually become a substantial market. The standards have been laid down, and there are various companies implementing them. Epok in Bethesda MD is among the most advanced in this regard. It has commercial product. And customers.

ID dataweb (actually, there are many synonyms for this emerging space—I’m partial to “federated resource sharing”) is an approach under which every data element in every database can conceivably be given a unique, fine-grained identifier—thanks to XRI, which is backward-compatible with the URI/URN naming scheme that has achieved ubiquity on the Web. Hmmm…I hadn’t thought of that…the World Wide Web was built on the “identity of things” (aka pages, scripts, etc.), leveraging URI, DNS, and IP.

Well, anyway, ID dataweb is an environment within which autonomous data domains can choose to selectively grant fine-grained data-access rights to external parties—and unilaterally rescind those rights. It leverages the identity federation and trust infrastructure being implemented everywhere through open standards such as WS-Security, SAML, Liberty Alliance, and others. It’s a standards-based flexible way of securely setting up and managing as-needed data-integration connections between autonomous organizations. Such as manufacturers, suppliers, distributors, and other participants in a supply chain. Or financial services firms engaging in dynamic partnering on equities underwritings. And so forth. Data integration/exchange/transfer is one of the principal tasks in any B2B collaborative-commerce partnering.

Stevie Wonder once sang “Don’t you worry ‘bout a thing, pretty momma, cuz I’ll be standing in the wings when you check it out.” Well, you gotta worry about “things.” People are going to be checking out those “things” right and left. Here’s an issue that the ID dataweb community must grapple with: As organizations expose/share/protect more of their fine-grained data resources through XRI/XDI, how are they going to manage the massive databases underlying the humongous “directories of things” that result?

Oh…that was unintended…Stevie had an album called “The Secret Life of Plants.” I cycled subconsciously from the Wonder of Motown to the wonder of ID dataweb, and then back again.

Oh well. There you have it. Stranger things have happened.