Things change, I'll resist the cliché.
During our three month journey it became apparent that we will want to reconsider the APIs and data structures used. We made a deliberate decision that our first release would be as close a representation to the original dataset as possible.
This was done for a number of reasons, most importantly because we wanted to learn. We wanted to learn what made a good dataset and what made a good API. We also wanted to learn how the showcase developers would use our APIs and use this to inform our next iteration of APIs.
So, the general principle we have come up with to handle versioning is this:
- The URL request can contain a version number.
- If you request a specific version of the API, we will (assuming it's still available) invoke it.
- If you don't specify an explicit version then you will get the current version.
- It will not be uncommon for there to be a influx version, i.e. one that is being worked on. At some time a in flux version may become stable and 'final' for that version number.
- We will aim to support the current and a number of previous versions.
- Version numbers will follow the x.y notation.
- We will provide information on the available versions of the API and recommendations for their use.
- The first public version of the API will be 0.1