I hate these things about OAuth2
- Inconsistent naming across implementations. The client secret, consumer secret, client key, client ID , token Url, service account, client email etc.. etc.. are all mixed up in a big melting pot of confusion. The Cloud console and the documentation give the same things different names, as do individual app write ups, and these of course all change across implementations.
- Redirection Urls. It can be an app that doesn’t support this, so of course this means that you have to jump through workarounds and hoops to make it work. Sometimes you can’t know in advance what the apparent script URL you are executing against is, and when you write shareable libraries for multiple people to use you have to carry all that complication with your library and pass it on to people who use them. It’s built around an expectation of interactivity that might not be part of the purpose application.
- When it gets to the important part of writeups of Oauth2 implementations, the author usually pushes you off to some general Oauth2 writeup, as if he never wants to speak about his terrible experience again. This means there are no good write ups of Oauth2 implementations around that can be adapted to specific problems
In this article I’ll try to lay out some OAuth2 scenarios, and avoid shying away from the nasty bits. Hopefully the excercise will also help me having to relearn from scratch every time I need to implement something with Oauth2, and with any luck, you’ll find it useful too.
It’s an authentication dance, designed to ensure that whoever is asking for access to a resource is a) allowed to b) likely to be who he says he is, accessing from an expected place It goes like this.
- You register an app, and get some credentials.
- You request access and go through a ‘user consent’ dialog. This let’s the user know that the application is about to access some protected resource, and gives the opportunity to allow it.
- If the user allows it, the application gets an access code as proof. How to get that is the first tricky part. A redirect URL that you need to have previously registered with your app receives this access code. This code gives the application the right to ask for a token to allow it to subsequently communicate with the resource it is after.
- With the access code, and some other secret information, the application can now ask for a token which it will use for subsequent requests
- An access token is returned and can be used up to some expiry time. In addition a refresh token is returned. Using this refresh token, a new access token can be asked for if the original one has expired, thus avoiding the entire dialog.
Google describes the process like this
As a process it’s fairly straightforward. However, implementation can be challenging.
For help and more information join our community, follow the blog, follow me on Twitter