This chance also introduces a "LocalMetadataCache"
object, and switches the repo querying to all work
on the notion of an "active data set" that avoids
any assumptions about the filesystem cache layout.
While it retains the single architecture assumption,
as while as the "only base + updates" repository
structure, it removes the hardcoding from the query
layer (it's still hardcoded in the download layer
for now)
This chance also introduces a "LocalMetadataCache"
object, and switches the repo querying to all work
on the notion of an "active data set" that avoids
any assumptions about the filesystem cache layout.
While it retains the single architecture assumption,
as while as the "only base + updates" repository
structure, it removes the hardcoding from the query
layer (it's still hardcoded in the download layer
for now)