future_plans.md

Plans for the future

Immediate future

• Improve documentation
• Add a way to shutdown gracefully when running as a daemon.
• Make certificate key offset procedurally generated, thus eliminating the need for it to be communicated
• Add support for pushing updates to devices.
• Add tests (Bad habits die hard)
• Externalize serializer templates so that templates may be customized
• Continue optimizing everything.
• Finish implementing query support in the discovery protocol.

Potential changes/improvements

Remove encryption in favor of a token system

• Tokens used to prevent tampering and prove identity
• Tokens based on same keygen function
• Add the request body's checksum to the keygen to prevent tampering
• Advantages:
  • Less computationally intensive
  • Less error-prone
  • Less likely to expose shared secret
• Disadvantages:
  • Communications made less private

Split the program into several smaller programs

• Advantages:
  • Easier to manage individual parts
  • Improved performance
• Disadvantages:
  • Larger memory footprint
  • More complex launch/shutdown procedure

Add a uuid field to the discovery protocol

• Do not respond to discovery requests that lack a valid uuid
• Advantages:
  • Prevent message amplification used in DDOS attacks by making it less convenient
• Disadvantages:
  • Additional complexity
  • Someone could obtain a valid uuid by eavesdropping on network traffic, nullifying this defense

Make the service model more host-centric

• Allow the device service model, as defined in userdevice.py to be reused for programming hosts
• Advantages:
  • Less work involved with setting up a host
• Disadvantages:
  • Makes the distinction between device and host even less clear

Long-term goals

• Make first-time setup easier by somehow eliminating the need for certificates.
• Extend language support for the API (port it to Rust, Java, etc.)
• Extend human language support (localize for German, French, etc.)