User Generated Content and the Social Graph by Chris Chandler.
Uses Twitter as a case study. Covers Gizzard and FlockDB, both of which were written in Scala.
Wants to coin the term “MoSQL! (More than SQL).”
A couple of points of interest to topic mappers.
Relationships maintained as forward and backward edges. That is:
“A follows B” and
“B is followed by A”
Twitter design decision: Never delete edges!
Curious if any of the current topic map implementations follow that strategy with regard to merging? Thinking that the act of creating a new item either references other existing item (in which case create new version of those items) or it is an entirely new item.
In either case, a subsequent call returns a set of matching items and if more than one, take the most recent one by timestamp.
As Chris says, disk space is cheap.
Works for Twitter.