[torstenvl]

The trivial non-tracking implementation would be checking if the domain of a requested page matches a certain list of known e-commerce sites. If so, then on load, query the other known e-commerce sites through a proxy to see if there's a cheaper price. No tracking necessary.

[jsnell]

That would be both highly non-trivial and non-private.

For the non-trivial part, the approach would be hell to scale to any significant number of e-commerce sites. You can't have the client connect to tens of thousands of sites on a page load. And how do you reliably find the matching product pages on the other sites? It's easy if you can crawl the sites, but extremely hard if you tried to do it with point lookups.

As for the privacy, who operates the proxy? Why is giving them both your IP address and a stream of all commercially interesting web pages you visit not a tracking concern?