rest/urls
URL Conventions
The Simply Restful plugin for Rails, which will soon become part of Rails Core, has gone with the following scheme:
/people /people/1
We recommend that other implementations follow the same conventions, since that is the first concrete example of such explicitly RESTful URLs in the wild.
ARCHIVE
There is nothing that says you must organise resources hierarchically, although many people prefer to do so. There are advantages to separating your containment hierarchies from your components. It makes it trivial to add new containers and the containers can accept search parameters.
Example URLs
You could organise your URIs like this:
- The second widget
http://example.org/widget/2
- A form describing how to create a widget
http://example.org/widget
- The list of the first N (say N=25) widgets
http://example.org/widgets
- The list of the next N widgets starting at 26
http://example.org/widgets/26
Example Actions
Then consider a sequence of hypertext style actions:
- Retrieve a list of widgets
GET http://example.org/widgets
- Retrieve details of the second one
GET http://example.org/widget/2
- Replace some of the details
POST http://example.org/widget/2
- Retrieve a form (resource parameter decription)
- or whatever you want to call it, for creating widgets
GET http://example.org/widget
- Create a new widget
POST http://example.org/widget <= LOCATION: //example.org/widget/123
- Get the new widget
GET http://example.org/widget/123
- Delete the Widget
DELETE http://example.org/widget/123
Notes
This organisation is similar to that of the rest-discuss html interface. But you cannot use PUT, POST and DELETE in that interface.
Using the http://example.org/widget URI for creating new widgets is similar to the Prototype design pattern while still adhering to the expected behavior of a POST, ie. create a subordinate resource.
Whatever you do, be sure to tell clients what structure they use, so they don't have to guess (which would violated HTTP rest/opacity.