Inspired by @killercup's Static FileZ this is a library intended for producing compressed and archived versions of static file trees, with different tradeoffs than most general purpose archive formats.
See Static FileZ: What and Why for some of the background
constraints. The major deviations from that design are around flexibility;
Oubliette is intended to support multiple compression algorithms, different
storage backends (including async ones such as doing Range requests to S3)
and be able to configure the builder to tradeoff between total archive size and
extraneous data needing to be discarded to read a specific file (if you can get
the archive small enough, then instead of doing a remote request to access data
it may be cheaper to do extra work while reading from it locally).