Improve caching layer for Java Client

A first implementation has been done here: JAVACLIENT-24.

Here is the all discussions around the cache layer for further purpose in the README cache layer part -> update it to fill the missing part.

Goals

If needed, for example on Android, we should be able to easily add caching logic.

How?

All caches should be accessible via a generated cache key defined by the request itself:

• headers
• base url
• endpoint used
• parameters
• body
• content type
• ...?

How many?

3 caches should be implemented:

• Raw Response Store : The server response is simply stored on the device so that it can be reused in case the server is unreachable OR to avoid too many frequent calls.
• Document Response Store: Store the unmarshalled response objects (here Documents) and updates.
• Document Transient Store bound with deferred calls queue: keeping changes of document.
• Deferred Calls Queue: The Create Update Delete operation will be stored locally and replayed when the server is available. Requests pure calls.

• Actions/Events

Scenarii

Pending questions: Invalidations

----> What would be a default timeout for each cache?

Potential rules offline:

• When listing documents, check the document transient store
• then check the document response store
• then check the server response

Synchronisation

• Should we apply those rules ?
• Should we use ETag And/Or If-Modified-Since with HEAD method ?

Potential Stores

Depending on client:

• "In memory" - guava for java
• "Database" - SQlite for Android
• Local storage for JS
• On disk for both
• Others?

Miscellaneous

• For the dirty properties of objects (like dirty properties of automation client for documents) - out of scope of caching