auth0-sandbox-ext
v1.0.0
Published
Webtask extensions of Auth0 Webtask
Downloads
6
Readme
Webtask extensions
This project implements extensibility points of the Auth0 Webtask runtime using Auth0 Webtasks themselves. It also provides a pattern of local development of webtasks which combines GitHub code management, local testability of webtasks, and webtask deployment.
JWT token revocation
One extensibility point of Auth0 Webtask runtime is support for JWT token revocation. JWT token revocation requires durable storage so that information about revoked tokens can be queried and existing tokens can be marked as revoked. This functionality of Auth0 Webtasks is externalized behind an HTTP API and implemented as a webtask itself, so that various implementations can take dependency on variety of storage technologies. Current implementation is using DynamoDB.
The HTTP API is as follows:
HTTP PUT /?jti={jti}&jwt={jwt}
The PUT endpoint revokes a JWT token identified with the jti
identifier. The jwt
represents the entire JWT token being revoked which is stored for ease of future auditing. There is no validation of whether jti
matches the identifier in the token itself. The API returns HTTP 200 on success, or an error code otherwise.
HTTP GET /?jti={jti}
The GET endpoint checks the revocation status of a JWT token with specific jti
value. HTTP 200 status code indicates the token is revoked (i.e. the information about that token is present in the revocation database). HTTP 404 status code indicates the token is not revoked (i.e. no information about that token was found in the revocation database). Other codes indicate error conditions in the revocation check itself.
Webtask code
Webtask code for DynamoDB revocation check is here. The URL pointing to the raw representiation of this code, which can be used when creating the webtask token, is https://raw.githubusercontent.com/auth0/auth0-sandbox-ext/master/webtask/revoke_dynamodb.js.
Local testing
This repository establishes a pattern of local testing and development of webtasks.
The mocha test framework is used to execute tests.
The package.json file lists all Node.js modules the webtask code has a dependency on within the devDependencies
section. These modules inclue a subset of modules available in the Auth0 Webtask cluster, as well as additional modules that may be used by the test code itself.
NOTE The .env file placed at the root of the repository on the development machine (and excluded from source versioning) must include all secret parameters required by the webtask code. These are the same parameters that at runtime will be included in an encrypted form in the webtask token.
The test code follows this pattern:
- The .env file is loaded and secret paramaters added to
process.env
. In this case they include AWS credentials and parameters to connect to a DynamoDB table. - The webtask code to be tested is loaded and compiled into a callable JavaScript function using
eval
- just like it will be in the Auth0 Webtask runtime. - The
context
,req
, andres
arguments of the webtask function are mocked within the test code itself. The secret parameters from the .env file are added tocontext.data
. Thereq
object is constructed on a per-test basis. Theres.{writeHead|write|end}
is mocked to capture the response data. - When the mock of
res.end
is called, a callback function is invoked that returns control to test code and allows it to perform validation checks.
With this setup, webtask code can be developed and tested locally with mocha
.
Building webtask tokens
tl;dr
Basic idea is to have npm build
(or we can enhance webtaskify to support this mode) which will create a webtask token for all webtasks in the repository. By default the command will set the webtask code URL to the GitHub raw URL of the webtask code (this can be automatically determined from the environment). It will also add parameters from the .env
file as encrypted parameters to the webtask token. With this mechanism in place building webtask tokens will be a one-command process.