I was skeptical that fulltext search could keep pace with the entirety of a fast-growing web, but it has. So maybe this latency problem will be conquered. But the first examples of linked-data traversal that I’m seeing — in Freebase and DbPedia — are pretty slow.

If the improvement is only incremental then no. But one of the things that stops us from issuing more specific queries today is that they return no results. In that case the difference between no results and some results is more than incremental!

It wasn’t, the focus was mainly on crafting offers vs. crafting statements of need. However the latter is precisely what Doc et al. mean by VRM.

Also consider that once you’re not expecting synchronous response, you open the possibility for human involvement. You might as well consider things like “yahoo answers,” “amazon mechanical turk,” etc… the output from an “asynchronous search” service.

We often joke that it would be nice to have a human filter out our spam. I can see a service based on micropayments and Amazon’s Mechanical Turk model backing a VRM-focused search for purchase of many items, particularly large ticket items. Making virtual assistants affordable/usable by Joe Six-Pack. The key will be making the micropayment “commission” go toward the “best search” not necessarily just another paid form of spam funded by the sellers.