The description given on their Google Code page seems a bit sensationalistic to me. The author has a bit more reasonable look at the features and tradeoffs here: <a href="http://sbtourist.blogspot.com/2009/12/terrastore-and-cap-theorem.html" rel="nofollow">http://sbtourist.blogspot.com/2009/12/terrastore-and-cap-the...</a>
It looks like the complement to CouchDB: it actually implements all the hard distributed stuff that CouchDB never delivered on, but doesn't implement any of the fancy stored procedure stuff that CouchDB actually pulled off.<p>If only I could have the chocolate <i>and</i> the peanut butter together...
I wonder how much of that he got almost-for-free just by using Terracotta?<p>Terracotta itself is in effect a transparent ACID object database where collections can be larger than RAM and objects can be accessed efficiently on one machine or shared and updated in implicit transactions.