It does seem like it would take a hefty computer to run it, and I bet it wont work with PPC either.
My iBook wont run Second Life well..... So how will I buy a new MBP if they implement it? D:
Huh?
No, the hardware requirements would be pretty minimal. You aren't rendering a 3D store in realtime (although it leaves that open as a possible implementation); you are simply getting information about which other customers are in the store at the same time as you.
This allows the buyer to see which areas of the store are the most popular (if you come into Store X looking at electronics and see that 90% of their customers are in there looking at and buying books you know where the store's specialties are) as well as which specific products are gathering interest (in the electronics area if there's a mob around Router XYZ then it's probably either a good deal or a good product; you might want to look into that instead of / alongside Router ABC).
The pitfalls, however, are:
* Trustworthiness. In the "untrusted store" scenario, the buyer can not be sure the store is not astro-turfing "interest" in products it has an overstock of (just as such stores will manipulate front page placement and "similar users" recommendations based on stock).
* Mob mentality. This enhances the "mob mentality", showing sudden upsurges in product interest based on nothing more than other people having shown an interest in the product. This leads to unpredictable stock (an item hums around selling 2 units per week and suddenly today you get 20 orders in ... not good!) You already see this as a result of mentions in popular blogs, but this interface would speed up the flash mob effect by a factor of 10 or more. For instance, in a recent case a popular programming blog expounded on the benefits of a particular rechargeable battery charger; 24 hours later the price on Amazon had gone up by $10 (25%), and the wait time for the charger (still listed as in stock) was 4-6 weeks. Imagine that happening in 5 minutes instead!
* Intrusiveness. There are a good number of us who would happily answer the question "why did you go from looking at Post Raisin Bran to Kellogs Raisin Bran?" However, there are an even better number who would not want to be bothered by such questions, and were this interface ubiquitous would have to "opt out" of annoying browsing justification at Every Single Store on the Planet. The only reasonable approach for a store wanting to retain business would be an "opt in" basis, which likely would not get many takers. Throw in the likelihood of question abuses (see next), and I don't think even I would opt in anywhere.
* Spam. Imagine shopping at a store and getting bombarded with messages like "I see you are looking at Widget X. Did you know you can get the same widget at Mel's Discount Widgets for $5 instead?" No regulation on these would be practical (so they could be openly dishonest and aimed at building distrust of not just one store but two).
Overall, it sounds like a nice idea (I know it was on my "next steps for web stores" list for a few years), but it has some serious non-technical drawbacks. And, of course, the technical hurdles (needing to keep accurate and realtime track of where users are, which isn't easily done in the timing-out-sessions general design pattern for stores used today) aren't anything to laugh at either. Maybe Apple can clear them all. This one patent isn't far enough along to raise hopes, though.