routist

Objectives

• Facilitate creating HTTP Servers
  • provide very simple setup for common cases
  • provide flexible low-level components for special cases
• Client code (eg handlers) are not HTTP-specific, so can be reused eg in APIs
  • allows for future extensions to other messaging systems, eg sockets, pubsub, etc
• Provide flexible authentication/authorisation system
  • authorisation is out-of-the-box
  • authentication requires some user code, eg checking usn/pwd combo against DB
• Provide flexible dispatch system

Core Concepts

• HttpServer: top-level export; makes it easy to create HTTP servers

  • provide a factory function not a class
  • must pass auth/eval details to ctor

• User: opaque to routist code but passed through/sessionised; implemented by user code (simple strings?)

  • HttpServer can store User in HTTP session to associate user with a remote endpoint over time

• request and response values, both of type Message

• Message:

  interface Message {
      from: User          // expend in future
      to: User            // expand in future, eg groups
      subject: string     // indicates resource + intent (verb)
                          // better name? topic, headline, title, directive, summary, command...
      content: Payload    // better name? payload, body
      timestamp: Date     // added automatically
  }

• sentinel/special messages, eg 'unauthorised' response

• Payload type for message content

  • has a type prop enum defined and supported by routist
  • all other props are type-specific

• authenticate/authorise/evaluate

  • non-http-specific, therefore app-level controls must be in headline/payload envelope, not in eg http headers/status
  • need a number of sentinels so HTTP responses can use proper status codes. Which codes?
    • 200 OK
    • 400 bad request / client error
    • 401 unauthorised
    • 403 forbidden
    • 404 not found
    • 500 server error
  • support other common things in payloads, eg:
    • resource ID (ie the URL pathname)
    • optional params (ie the querystring)
    • username/password (for login route)
    • direct content with MIME type
    • JSON data
    • file upload/download
    • errors
    • paged data (both request and response)
  • not distinct steps/phases from routist perspective
    • all included in a single callback/option to HttpServer
    • most likely supplied by user as a decorated class (instance)
      • class props + values become evaluate multimethod predicates + handlers
      • authorise details given as property decorators
      • authenticate done through handlers (eg meta handlers)