To Polycy or not to Policy #1580

rkoepferl created 3 years ago

I have read and tried a lot to establish resource based access policies. Basically the classic: Only ever the creation user may access a resource (for simplicity) In order to judge whether the current user can access a resouce, we have to get the Entity, first - hence enter the handler. So [Authoirize]-Attributse are not useful.

Keeping an eye on https://docs.microsoft.com/en-gb/aspnet/core/security/authorization/resourcebased?view=aspnetcore-5.0 I came up with this:

        // create a policy based on claim audience
        context.Services.AddAuthorization(options =>
        {
            options.AddPolicy("CallRecordingAccessPolicy", policy => policy.Requirements.Add(new CallRecordingPolicy()));
        });

        context.Services.AddSingleton&lt;IAuthorizationHandler, CallRecordingAuthorizationHandler&gt;();

If you're creating a bug/problem report, please include followings:

CallRecordingPolicy being an empty class; "CallRecordingAccessPolicy" being never used anywhere else. The implementation of the Handler is somewhat evident and the AppService is doing something like that:

    public virtual async Task&lt;CallRecordingDto&gt; GetAsync(Guid id)
    {
        CallRecording recording = await _callRecordingRepository.GetAsync(id);
        AuthorizationResult result = await AuthorizationService.AuthorizeAsync(resource: recording, requirement: new CallRecordingPolicy());

        if (!result.Succeeded)

Essentially I use a lot of infrastructure with no real benefit. Also this is not right available in the DI container in the UnitTest. Testing is a bit cumbersome. And in fact I could just declare some other arbitrary class, hook it up to the DI-Container and use it as so:

public virtual async Task<CallRecordingDto> GetAsync(Guid id)
        {
            CallRecording recording = await _callRecordingRepository.GetAsync(id);
            bool result = await _callRecordingChecker.CanAccess( recording _currentUser);

            if (!result)

public class CallRecordingAuthorizationHandler : ITransientDependency
{
public Task<bool> CanAccess(CallRecording r, ICurrentUser cu)

So, why should I use the first approach over this second approach?

Advantage:

Simple, thats waht needed
easier to use in UnitTests
more Straight forward
automatically in the DI container

2 Answer(s)