Uniform interface

At the core of REST are resources, and resources are identified using Uniform Resource Identifiers (URIs). Conceptually, resources are separate from their representation (that is, the format in which they are provided to clients). REST does not mandate any specific format, but typically includes XML and JSON (which will be discussed in Chapter 4, Data Representation).

In addition, resource representations are self-descriptive. In more concrete terms, this means that sufficient information must be returned for the successful processing of responses.

Another distinctive property of REST is that clients interact entirely through hypermedia, which is dynamically provided by the application servers. Apart from endpoints, clients need no prior knowledge of how to interact with a RESTful service. This constraint is referred to as Hypermedia as the Engine of Application State (HATEOAS).

Client-Server

The client-server model that REST embraces enables the separation of client concerns, such as user interaction or user state management, from that of server concerns such as data storage and scalability.

This decoupling ensures that, provided an interface that is agreed upon, the development of client and server can be done independently. It also helps reduce complexity and improve the effectiveness of performance tuning.

Stateless

REST advocates statelessness. No client state is stored on the server. All the information needed to perform operations is contained in the requests (as part of the URL, request body, or as HTTP headers).

Cacheable

RESTful web services must provide caching capabilities. Servers can indicate how and for how long to cache responses. Clients can use cached responses instead of contacting the server.

Since REST typically leverages HTTP, it inherits all the caching properties that HTTP offers.

Layered system

Given the style of communication between clients and servers, clients are not aware of which specific server they are interacting with. This property allows the introduction of intermediate servers that can, for example, handle security or offer load-balancing capabilities. These architectural concepts are discussed in more detail in Chapter 10, Scaling a RESTful Web Service.

Code on demand

Even though it's part of the REST architecture, this principal is optional. Servers can temporarily extend the functionality of clients by transferring executable code. For example, JavaScript can be provided to web-based clients to customize functionality.

For a service to be considered RESTful, it should abide by the preceding principles.