for anyone interested, but without the time, he focuses on monotonic processes (things that only get bigger, in a sense, which are the kinds of things that become eventually consistent - think of the max of some values, or the set of all instances in a collection that is accumulated).<p>then he expands on a language that uses datalog (without deletion, presumably) to implement this (which makes an awful lot of sense, as - iirc - datalog systems were implemented by constructing all possible solutions, which looks like it's going to be monotonic as you accumulate data and functions...)<p>it really is a good talk if you know nothing about all this (as i didn't).
Hi folks. Related presentation: <a href="http://research.microsoft.com/apps/video/default.aspx?id=153540" rel="nofollow">http://research.microsoft.com/apps/video/default.aspx?id=153...</a>
Talk is about 36 minutes (then questions). The theoretical part at the beginning is the best, and then seeing the actual implementation for Bloom was very informative. I do think this will highly influence my solutions for these kinds of problems, and this talk is hitting the nail on the head for the kinds of problems you have to deal with in eventually consistent systems.<p>Great stuff!