List of things that need to be implemented

Short-term

document
- (better) README
- the design
- user-visible UI etc
measure, improve test code coverage
more metrics endpoints - think on what is really needed
- perhaps backup/snapshot/restore sizes and file counts? copy from *hoard?

Eventually

Cassandra partial restore support; notably, we should be able to recover single nodes (perhaps in this case, if version matches, restore both system keyspace AND user keyspaces, and avoid the create steps in full restore)
Cassandra specific table restore support; we should then pick more carefully where exactly we are restoring things:
- get table ids for freshly created tables (during restore when Cassandra is up from system_schema.tables)
- pass the keyspace+table+table id tuples to the actual restore Cassandra step
  - clear the directory
  - move files to it
package (or have someone do it?) this for distros
page the result from operations (as it is, some of the stuff may take e.g. 30 seconds to get across in our prod cluster)
push this to PIP
use result-url instead of polling for somewhat faster results for CLI
we should have some sort of packfile format for small files - right now, the embed-in-manifest kind of works, but is not pretty, and doesn't work for still trivial sized files

Maybe not - known design choice for now

in m3db placement plan, we do not rewrite port number
it would be possible to split (large) BackupManifest to e.g. BackupSummary which would be used by list backups endpoint; in practise, the results should be cached locally so cost of downloading (even largish) files once is not prohibitive
permissions / file modes and directories/links in general are ignored. if this is actually used for general cluster backup, actually storing those would be nice too and not too much effort. however, it is not really focus of the project for now so left not done.