Representational State Transfer (REST) is a software design architecture used to communicate state between two systems that has no current IndieWeb adoption, likely due to its fundamental incompatibility with static sites. REST is typically associated with web services and applications because the HTTP protocol contains a set of generic verbs (GET, POST, DELETE, etc.) that can be applied to an arbitrary set of nouns (URIs).
Why use REST? (to be filled in by someone who thinks it's a good idea)
How to use REST on your site (to be filled in by someone who is selfdogfooding REST on their own site).
No current IndieWeb examples of "REST" APIs on people's personal sites.
Applications are said to be RESTful if they are designed with the following constraints:
- Separate concerns between the client and the server. The client is not concerned with data storage and the server is not concerned with user interfaces or user state.
- Maintain a stateless server. No client context should be stored on the server between requests.
- Allow responses to be cacheable in order to reduce client-server interactions.
- Having a client talk directly to a server or routing the request to an intermediary server should have no effect on the response.
- A uniform interface that allows clients to identify resources, manipulate them, and interpret responses.
Not quite REST
There are APIs that often either claim to be REST but do not conform to the above constraints, or claim some similarity to REST by using the following terms:
Pure REST fails with Static Sites
Pure REST requires that individual URL support all the various HTTP verbs directly (i.e. beyond GET, explicitly PUT, POST, DELETE) and static site servers typically do not support them, hence static.
Since static sites are a common IndieWeb hosting use-case, this means that pure REST APIs are a showstopper and cannot be used for IndieWeb interop.
Instead, rel discovery based alternatives like Micropub and PuSH are far more practical, allowing for delegation of various operations including editing.
Alternatives to REST
Remote Procedure Calls
In an RPC architecture, a client sends a server a set of parameters so that the server may then execute a set of procedures. In the web application context, RPC implementations typically target a single endpoint and use POST to send parameters to that endpoint, in essence using HTTP as a tunnel for a custom protocol. As a result, there are a large number of largely incompatible RPC protocols. Examples include XML-RPC, JSON-RPC, SOAP, etc.
Since clients need prior knowledge of available endpoints and acceptable parameters before beginning an interaction with the server, clients and servers are typically tightly coupled in RPC architectures.