Support multiple security contexts within an application

Most RBAC systems provide a simple permission model to answer permissions. For example, is the user authorized to POST a comment. In an application that supports multiple security contexts the relevant context is an additional parameter required to determine if the user is authorised; that is is the user authorized to POST a comment to this project vs another project. A user may be authorized to POST a comment to one project, but not another.

Support for security contexts at the query level

If every object has a security context property it’s possible to always secure the queries executed for a user request by adding the user’s authorized security contexts as an “IN” condition restricting the scope of the query.

Support assignment of roles within a security context

In most RBAC systems a user is assigned a role that is applicable across the entire application. For example, a user may have a Commenter role granting the user the permission to post comments anywhere comments may be posted within the application. An application that supports the assignment of a role within the scope of a specific security context can restrict the assignment of the Commenter role to specific areas of an application. For example, a user may be assigned the Commenter role in Project 1 but not Project 2.

Support an extensible mechanism for conditional policy expressions

Another limitation of most RBAC systems is the simplicity of their permissions - you are either authorized to perform an action or not. Supporting an extensible mechanism for conditional policy expression means that a permission may be conditionally granted dependant upon a variety of factors. For example, you may be granted the permission to comment within a project only when a target resource within the scope of that project is “open for community comments” vs “open for drafter comments”.

Support JSON representation of total permissions granted

If a user has permission to perform an action, such as update a resource, it should be possible for a client to determine that this is the case and be able to decide whether to disclose editing capabilty for the resource to the user. That is, if I’m not authorised to edit a resource the UI shouldn’t offer me the opportunity to edit the resource.

Providing a JSON representation of the user’s total permissions, and a Javascript implementation of the permissions policy engine, can provide a much richer user experience.

uv add django-scoped-rbac   # or: pip install django-scoped-rbac

# myapp/workflow.py
def is_allowed_transition(current: str, proposed: str) -> bool:
    ...

# settings.py
SCOPED_RBAC = {
    "CEL_FUNCTIONS": {
        "is_allowed_transition": "myapp.workflow.is_allowed_transition",
    },
}

uv sync                              # create .venv with the dev dependencies
uv run pytest                        # run the tests with coverage
uv run ruff check && uv run ruff format --check
uv run --group docs make -C docs html   # build the docs into build/sphinx/html