The biggest resulting change is that the consent page always appears
after logging in.
Some minor fixes to be closer to the spec are included
together with some minor structural refactors.
* Solution works but tests don't
* refactor(FetchUtil): use arrayifyStream
* refactor(FetchUtil): split fetchDataset into 2 separate functions
* style(FetchUtil): onelining instead of declaring new local var
* test: trying to mock rdfDereferencer
* refactor: promise can't have async function as arg
* test(FetchUtil): pass Quad array to mockDereference instead
* test: all tests should pass now and coverage is back to 100%
* style: comment typo
* chore: make package.json and package-lock.json compatible with main
* chore: fix package.json double entries
* chore: updated package.json to be alfabetical again
* refactor(AgentGroupAccessChecker): Remove converter from contructor and config
* refactor(TokenOwnerShipValidator): Remove converter from constructor and config
* refactor(FetchUtil): Return BadRequestHttpError instead of generic Error
* test(FetchUtil): return Response object instead of mocking fetch
* style: typos and newlines
This is a revert of a previous change
but is now possible due to the use of JSON bodies.
This does mean JavaScript is required in the HTML page,
but that will be required for future changes anyway.
This allows easier reuse of certain reoccurring behaviours,
such as authorization.
The AuthenticatedLdpHandler is no longer required
since it is a combination of parsing and authorization.
This did require a small change to the OperationHandler interface.
Account settings are stored using the WebID as key.
Reason for using the WebID is that this allows faster access to the settings
in authenticated requests.
A consequence of this is that passwords are now always required during registration,
and that there can only be 1 account per WebID.
The IDP behaviour has been changed to move all error related knowledge
to the IdentityProviderHttpHandler instead of managing it
in the Interactionhandlers.
Each IDP class using storage now has a different storage.
This way those classes don't have to worry about clashing keys anymore.
All internal storage is now in the /.internal/ container,
thereby making it easier to take the location of the internal data into account:
only 1 path needs to be blocked and a regex router handling internal data
differently only has to match 1 path as well.
