This is an early days experiment to implement a document-oriented Remote Content Access HTTP API for Sling content.
Document-oriented means that the API exposes higher level objects than Sling resources, typically documents which represent website pages and similar content objects.
The API is meant to be fully discoverable, starting at the root should be sufficient for you to discover all the content and how to use it.
It is meant to be friendly to both humans and machines, which is a challenge. The well-defined scope of the API helps with that.
The API itself expose read-only content, but it includes links to a command processor which will be implemented (separately, later) to process content modifications and other commands. That processor will use a CQRS-style Command Pattern, where the read channel represented by this API is distinct from the write channel (the Command Processor) and where things are not guaranteed to happen synchronously.
Build and start this with
mvn clean install exec:java
And open the root page at
http://localhost:8080
After that...well, the API is supposed to be discoverable so you should find your way!
To activate debugging, use the standard MAVEN_OPTS - the Java code is started directly in the Maven process.
The /content/articles subtree is where the most interesting content is, for now.
For this use case, we want the API to be easy to consume both by the developers who learn and discover it, and by client-side rendering code.
The client-side rendering code will usually not be provided by Sling for this use case, but it could, using the mechanism described below for edge-side rendering.
For this use case we probably need to deliver both the content and the rendering templates from Sling. We might use the current template resolution mechanism and include the URLs of suggested templates, keyed by extension and selectors, with the content in order to simplify the edge-side rendering code.
The test content uses com.adobe.aem.guides:aem-guides-wknd.ui.content.sample which is MIT licensed. Minimal “fake” JCR nodetype definitions are used to allow this content to load, as we don't really care about the details of these node types besides their names.