From the webpage:
GraphPack is a network of autonomous services that manage graph structures. Each node in those graphs may refer to a node in another service, effectively forming a distributed graph. GraphPack deals with the processing of such decentralized graphs. GraphPack supports its own traverse/query language (inspired by neo4j::cypher) that can executed as transparently distributed traverses.
Amit Portnoy wrote about GraphPack on the neo4j mailing list:
The prototype, called GraphPack, has a very lite design, actual persistence and communication aspects can be easily pluged-in by injection (using Guice).
GraphPack enables transperantly distributed traverses (in a decentralized graph), which can be specified by Cypher-inspired traverse specification language.
That is, clients of a GraphPack service (which may have a graph the refer to other nodes in other GraphPack services) can write a cypher-like expressions and simply receive a result, while the actual implementation may make many remote communication steps. This is done by deriving a new traverse specification in every edge a long specified paths and sending this derived specification to the next node (in other words, the computation moves along the nodes that are matched by the traverse specification).
Sounds like there should be lessons here for distributed topic maps. Yes?