One option is to implement the version number as the context root of the app. Then you just leave v1 app running while you code and release v2 app. This allows

One option is to implement the version number as the context root of the app. Then you just leave v1 app running while you code and release v2 app. This allows complete isolation, but may be harder to manage if you have architecture/security changes and want to upgrade v1. Another approach I have seen is just providing backwards compatibility where it matters or where there are changes. This would be similar to how a library works. You try to limit things that need to do if-def on version number. Usually only 2 versions are maintained simultaneously. v2 still supports the requests of v1, but when v3 is released v1 features are gone. I am sure there are other methods. Edit: https://www.google.com/search?q=stackoverflow+api+versioning https://www.google.com/search?q=stackoverflow+api+versioning