-3

Okay, to start off I am new to the world of working with servers and RESTful clients. What I am currently doing is working on a RESTful client api and I realized as I was chugging along that there were times when the api would completely fail due to errors in different parts of the code. I am working with Django on Amazon EC2 servers.

The point of my question is: what is good practice in setting up a way to work on the api without constantly causing failures? I've thought about having a variable that directs flow through the client based on whether the request is a test or not, but that still does not solve the issue of larger problems that may occur.

Advice, leads and reading material is appreciated. I've looked around and asked advice of some people, but I am still fairly lost. At this point, the easiest option seems to be to make a complete duplicate environment for testing (so a whole new server) and only push the changes when the development is stable. This just seems really inefficient though.

bmargulies
  • 97,814
  • 39
  • 186
  • 310
Kinetic Stack
  • 788
  • 12
  • 33
  • Well actually the practice of having multiple environments (often dev/staging/production)is probably the most common and the most safe one. – Manu Clementz Jun 12 '13 at 18:42
  • But how do I set that up? Should I create and entirely new server instance with the same configurations to do the development on? I've seen some say that local works, others say that having another server instance purely for testing is the way to go, and I feel a little lost in it all. – Kinetic Stack Jun 12 '13 at 18:47
  • 1
    It depends on your environment / stack, but there are a lot of tools to manage development environments. I use [Vagrant](http://www.vagrantup.com/) a lot for dev boxes. – Manu Clementz Jun 12 '13 at 18:50
  • I am looking into that, thank you for the direction. – Kinetic Stack Jun 13 '13 at 15:20

1 Answers1

2

A restful system should use an entry point URI and discover the other URIs from within returned representations. Also, the URI structure should be opaque to clients so it should be possible to define two URIs like,

http://acme.com/prod/api

and

http://acme.com/dev/api

The client should accept one of these URIs and as long as the URIs that are returned in the representations by the server respect the target environment then everything should just work.

The idea of hypermedia and opaque URIs is designed to make clients ignorant of any particular server implementation therefore what you are trying achieve becomes very easy.

Darrel Miller
  • 139,164
  • 32
  • 194
  • 243
  • Okay, so this approach definitely makes sense, but then what is the best way when modifying shared functionality? Would all functionality just be duplicated for each endpoint and one of the duplicates be the 'dev' functionality? To be clear, say there is a shared Python script that serves a purpose, then would it be best to make a copy of it for the dev calls and then push it to the prod side? Based on what I can find online this seems like an approach, but then in a Python environment it is difficult to control things that may break it overall. Is this the purpose of virtualenv? – Kinetic Stack Jun 13 '13 at 15:19