> I am researching how to set up a schema for querying a set of tags
> associated with an object.
I would be interested in hearing your conclusions.
I am currently researching in a similar direction.
We have streaming replication where the slaves are used for data mining, storing currently about 100m objects and
expectedto grow in direction of 1-2b or more.
While part of the data is stored in a classic RDBM model, there are "soft" properties, for which we are looking for
"soft"and in the same time, if possible, searchable/indexable storage.
So far I have mostly looked into hstore and JSON, but arrays are an interesting option, as well.
The only conclusion which we have reached so far is that we should divorce the core storage from search optimization
andfrom versioning.
T.E.L.