I've been reading up on the Learning Registry recently, which is a
distributed system for learning resource metadata that includes both
"regular" descriptive data and usage data ("paradata") such as
reviews, downloads and ratings.
E.g., the following gets information about a resource (something about
KC-135 tanker aircraft) and also paradata that asserts 73 views and 11
downloads:
The interesting thing is that multiple "repositories" can publish this
data to a LR node, and that LR nodes can share this data in a
network.
So in our case we might have a network of LR nodes for widget
metadata, which is contributed to by ROLE, ITEC and OU widget stores,
meaning we could pool the usage metadata to improve recommender
performance.
Some more info on the LR node being set up in the UK here:
http://blogs.cetis.ac.uk/lmc/2011/11/07/jisc-learningreg-node/
S