Currently, someone setting up a local development server only has a choice between importing the full data set with millions of artists, releases and other entities or no data at all (except for a few bits like release status that are required for the server to run properly, and therefore inserted by InitDb.pl --clean). We should provide another option: a reasonable subset of the data that contains examples of "everything" (all entity types, artist credits, most relationship types, probably all countries/languages/scripts) but is still much smaller than the full database.

It would be enough to dump this subset only from time to time (at least after each schema change, but probably a bit more often); still, the process should be mostly automated or it will not be done. Bitmap proposed choosing a number of "representative" artists, then including everything related to them in the dump.