Unintentional collisions wouldn't happen: each manufacturer gets a unique ID, and in turn gives each product type, and each instantiation of each product of that type (e.g., this particular 12 oz. can of Coke) a unique serial. Only that particular manufacturer gets to assign codes with its manufacturer ID (or, as EPCglobal calls it, an EPC Manager Number) in it, and presumably won't screw up when it assigns its own IDs to its own products.
Why not? The idea with the ONS is that someone (VeriSign, per the contract that EPCglobal let) will run a fairly small (and replicated by others) root service to say, "If you want to know about EPC=XXXXXX..., you need to look over there," and give a pointer to PepsiCo. At PepsiCo (or some agent of PepsiCo's choosing, say IBM, or GXS, or whomever), there'll be services to further parse the request, and direct it to an appropriate target. PepsiCo could choose to construct a single huge database with entries for every tag (associated with every product) it creates, though it need not... that might be broken up among various bottling units, etc... we need to think of "EPC space" as a vast, federated landcape of services.
The elegance of the EPC is that it parses into parts: a part will say, "This EPC was assigned by PepsiCo," a part will say, "It corresponds to this PepsiCo product," and a (fairly large) part will say, "For this PepsiCo product, this particular EPC represents this specific unit."
Airline seats are real, but they're not "bought" or "sold" (or, in this case, rented or scheduled) with the physical good on hand. What I meant in the article was that there may be a utility to adopting the EPC as a standard reference to saleable things, whether or not there's a physical good to stick an RFID on.
The example of Amazon.com and its Web services API is somewhat relevant to this... Amazon has made it possible for its affiliates to fetch and display book information, if they make a request based on those existing standards, the UPC, EAN or ISBN. It's a shame I can't currently use a UPC universally (and this is what the CueCat folks were aiming at) to fetch product information from the authoritative sources, but the Object Name Service should make that possible for EPCs (which include UPCs and other legacy codes within them). That will dramatically simplify Internet commerce, I think, and expanding EPCs to other areas--including to airline seats--would improve things there as well.
Unintentional collisions wouldn't happen: each manufacturer gets a unique ID, and in turn gives each product type, and each instantiation of each product of that type (e.g., this particular 12 oz. can of Coke) a unique serial. Only that particular manufacturer gets to assign codes with its manufacturer ID (or, as EPCglobal calls it, an EPC Manager Number) in it, and presumably won't screw up when it assigns its own IDs to its own products.
Why not? The idea with the ONS is that someone (VeriSign, per the contract that EPCglobal let) will run a fairly small (and replicated by others) root service to say, "If you want to know about EPC=XXXXXX..., you need to look over there," and give a pointer to PepsiCo. At PepsiCo (or some agent of PepsiCo's choosing, say IBM, or GXS, or whomever), there'll be services to further parse the request, and direct it to an appropriate target. PepsiCo could choose to construct a single huge database with entries for every tag (associated with every product) it creates, though it need not... that might be broken up among various bottling units, etc... we need to think of "EPC space" as a vast, federated landcape of services.
The elegance of the EPC is that it parses into parts: a part will say, "This EPC was assigned by PepsiCo," a part will say, "It corresponds to this PepsiCo product," and a (fairly large) part will say, "For this PepsiCo product, this particular EPC represents this specific unit."
Airline seats are real, but they're not "bought" or "sold" (or, in this case, rented or scheduled) with the physical good on hand. What I meant in the article was that there may be a utility to adopting the EPC as a standard reference to saleable things, whether or not there's a physical good to stick an RFID on.
The example of Amazon.com and its Web services API is somewhat relevant to this... Amazon has made it possible for its affiliates to fetch and display book information, if they make a request based on those existing standards, the UPC, EAN or ISBN. It's a shame I can't currently use a UPC universally (and this is what the CueCat folks were aiming at) to fetch product information from the authoritative sources, but the Object Name Service should make that possible for EPCs (which include UPCs and other legacy codes within them). That will dramatically simplify Internet commerce, I think, and expanding EPCs to other areas--including to airline seats--would improve things there as well.