I am sorry but this does not explains map reduce in the way it is currently used, wikipedia does a much better job http://en.wikipedia.org/wiki/MapReduce

Amazing how 44 people upvoted an incorrect example.

Why do you think voting something up is corroboration? People read something that was internally consistent and engagingly presented. They learned something from it, more than any inaccuracies in the details. I think that merits an upvote.

People will upvote far more egregiously false content if it's well presented and seems to be internally consistent. They thought they learned something from it. They weren't voting as experts on the material. There's nothing wrong with this. The system is working as desired.

Voting by correctness easily slips down the slope to voting by agreement, and that leads to a less interesting, more echo-chamber-y community. Vote up interesting stories even if they're inaccurate.

I guess Ayende first heard about MapReduce in CouchDB context, while most people talk about MapReduce in Google/Hadoop context.

This all what Erlang programmers need to know about M/R:

    -spec split(term()) -> [{k1(),v1()}]. 
    -spec map(k1(),v1()) -> [{k2(),v2()}]. 
    -spec reduce(k2(),[v2()]) -> [v3()].