id: 'idea--feed-reader'
content: |
  There's no reason for a feed reader to be a single program. Instead,
  one could build an architecture in which each component is a distinct
  program that doesn't need to care about the details of the others.
  For now, I'm going to refer to this system as _lektor_.

  The `lektordir` format is inspired by the `maildir` format for email
  but modified for feed readers. It similarly includes three
  subdirectories: `tmp`, `new`, and `cur`. File name generation works
  similarly.

  However, instead of writing email files, we need a standard way of
  describing a feed entry. Therefore, instead of writing _files_ to
  a `lektordir`, update daemons will write _directories_. A given
  feed entry consists of a directory containing at least four
  files: `title`, `url`, `id`, and `content`. Other optional files can
  exist, including `author`, `subtitle`, `summary`, and so forth.

  With this system, adding a new feed type—such as a Twitter feed or
  an ActivityStream serialization—becomes a simple matter, and an
  _lektor_ front-end can be created without having to worry about how
  that data was delivered there in the first place.
related: []