ETag support

[tdammers]

For template-based responses, this amounts to hashing the entire response body.

For static responses and raw JSON responses, we can suffice with hashing backend data, but the rendering overhead for those is minimal anyway, so it's probably simpler to just always hash the full response body.

Performance wise, it would be great to be able to hook into backends to directly query them for changed data (which is possible, in principle, for the literal, HTTP, and file backends); unfortunately the load() template function makes it impossible to infer which backends will be hit during a request, so this is only going to be useful for rules that don't involve templates (raw JSON and static).

Relevant discussion in #17.

[andrewufrank]

thank you for your patience and the clarification of facts - i learn more about http than I ever wanted.

My question arose in a combination of

• handling errors on my side (not saving a file after editing, duplication of files with the same name etc. --- all the stuff occuring when one does not know much and tries to learn)
• not undertanding when a restart is required and when not.
  I hope you can add a sentence of two to clarify the later - i may send you a pull request with a suggestion.