The mytoken service is heavily connected to OpenID Connect (OIDC). That is why we want to explain the basics here. If you are familiar with OIDC skip this page and go directly to The Mytoken.
OIDC gives users a way to log in to different sites using a single account at an identity provider. If users want to access another site, they are redirected to the OpenID provider where they log-in and then are redirected to the target site. The target site obtains all relevant information from the OpenID provider through different tokens.
OpenID Connect Tokens¶
OIDC knows different tokens that all have different purposes. In the following we will briefly describe these tokens.
Access Tokens are short-lived credentials that are used as authorization for accessing protected resources. They can be passed around.
Access tokens are the most important tokens in our context. They are used for authorization at a resource server. This means that a client that wants access to a resource - e.g. an image on a cloud storage - presents an access token to the cloud storage server in order to get the image. The cloud storage server will verify the access token and if everything checks out return the image.
Access tokens can be used for different operations, depending on the resource server and the scope of the token.
So an access token can be used to actually do something. However, because anyone with the access token can do that something, the lifetime of such an access token is strongly limited, usually in the range of minutes up to an hour. This means, that after that time the access token can no longer be used, and a new access token is required.
Refresh Tokens are long-lived credentials a client can use to obtain additional Access Tokens. Refresh Tokens are powerful, must be kept secret, and cannot be passed around.
Sometimes it is necessary for a client to do operations for a longer period than the lifetime of a single access token, e.g. if a user wants to stay logged-in in a web application. In such situations the user does not want to re-log-in each time an access token expired, so another menchaism is needed. (This is even more relevant in use cases where the user is not present anymore, and the application should do things on his behalf.)
Therefore, the client requests a refresh token from the OpenID provider when the user logs-in in the first place. Now the client does not only receive an access token from the OpenID provider but also a refresh_token. The refresh token can then be used to obtain a new valid access token when the current one expired. This way the client can always have a valid access token.
For this to work the refresh token must be long-lived. This is the case, often the refresh token does not expire at all. Because the refresh token is so powerful, since it allows a client to obtain access tokens for a generally unlimited time, these refresh tokens are bound to the client that requested them. This means that - other than access tokens - refresh tokens cannot be passed around.
ID Tokens encode user information, that can be used by the client, e.g. the webpage gets the user's email address from the provider.
mytoken context ID Tokens are not used.
scope of an Access Tokens defines what can be done with that token.
As noted above, access tokens can be used to do different things. What can be done with a particular access token is
controlled through a mechanism called
scope. An access token has one or more scopes attached to it.
There are some well-known, predefined scopes, like
openid scope gives access to an identifier string, the
profile scope access to basic profile information, and
An HPC cluster for example could require an access token with the
compute.create scope in order to create / start
compute job, while the
compute.cancel scope is needed to cancel the job.
audience of an Access Tokens defines where it can be used.
While scopes can be used to limit what can be done with an access token,
audience can be used to limit where an
access token can be used. Therefore, one or more audiences are encoded in the token.
When a resource server obtains a request with an access token, the server checks if the access token can be used at this
resource, i.e. it checks if it is listed as an audience.
The exact value of an audience claim is not defined. It is recommended to use the url of the resource server, but it could also be something more general. The important aspect is that the resource server accepts the audience value.
A cloud storage server could decide to accept the following audiences:
Note that currently only very few providers support requesting access tokens with specific audience values.