DB feature spec

Feature requirements for DB platform

Must haves

• ability to add new columns at any point (strongly suggests NoSQL)
• good JSON support
• authentication
• permissions (i.e., authorization)
• Python bindings actively under development

Preferred

• ability to sync local <-> master
• web interface with CRUD operations
• git-like history (i.e. easy to roll back or diff commit1...commit2)
• ability to preview changes before syncing (I (james) think this is a must have, or at least to build the tooling to show this before updating the DB)

Open questions

• REST API layer for CRUD operations? Or done through direct DB connections (via scripts and web client)?

Software packages

Databases

• RethinkDB ---
  • NoSQL database
  • robust, awesome api, friendly query language
  • backup / restore exists
  • doesn't sync, but thinker adds sync functionality to rethink
  • Provides database- and table-level permissions
  • Officially-supported Python bindings
  • nice guide to implementing a REST API with Rethink + Express
• CouchDB
  • NoSQL database
  • main feature is sync, probably the most mature platform with sync
  • built-in web interface (fauxton)
  • SQL-style declarative API
  • No well-maintained Python bindings
  • Limited control over permissions: Supports authorization classes for "members" and "admins" both of which allow read/write on any document in the database
• MongoDB
  • NoSQL database
  • SQL-style declarative API
  • sync by dump/import or by cloning
  • web interfaces through community projects like mongo-express
  • built-in Python bindings (PyMongo)
  • Database- and collection-level permissions implemented through user roles with ability to restrict users to specific set of commands on a specific collection
  • Document validation to enforce a minimal schema
• Dat --- distributed datastore, does distributed sync really well, low-level (HTTP rather than HTML)
• noms --- distributed database, higher level than dat, deals with structured data, syncs database base instances, importantly, query system has not yet been built / is experimental
• gun --- distributed database, similar in spirit to noms, great looking API
• datomic --- distributed, transactional, immutable, queryable, Would run on DynamoDB. Written in Clojure. Datalog query language (may be a steep learning curve). Git like history. Free for open source, else $5000/year. Flexible schema. Probably too advanced for us.
• chado

Backend / API

• Eve --- Python REST API framework

Other useful pieces

• Passport --- really nice looking library for managing authentication
• Horizon --- JavaScript backend, access RethinkDB database from the browser, manages sessions, auth, React-friendly