Essentially they're creating dedicated circuits between two endpoints to eliminate store-and-forward delay at backbone routers.

I think the title overdoes the impact of this: this isn't the kind of technology that is going to hook end systems to the core of the Internet, and this approach prevents routing from being done at the network level. I could imagine some kind of overlay routing scheme making use of this to connect points near the edge of the network. Or, as the article states, between big datacenters.

Sort of. Optical channel routers already exist. (I think they are even commercially available from companies like Infinera.) This work appears to address management protocols that rapidly redirect optical channels based on network traffic needs.

I am suspicious of the touted throughput increase, at least for personal applications. The public Internet tends to be dominated by ad hoc video and audio loads, which are fairly predictable in aggregate. There probably are some commercial users (like TV networks) who have bursty 100+ Gbps loads that could benefit, but most of them will want to purchase dedicated links.

totally unrelated, but when I clicked this link, the colors/layout made me think it was digg. If anyone in charge of the site sees this, something really should be done about it. Digg is really a terrible thing to be associated with.