8:23 a.m.
Hello all,
...more specifically people that use keycloak authorization services.
While dealing with RTPs (without permission tickets) for both user and centrally managed
resources we encountered an inconsistent behavior and would like to know if it is
considered a bug or works as intended (and why)
The story:
When a resource is owned by the resource provider (a client), you can get a RTP by
providing either the resource id (uuid) or the resource name in the
"permissions" parameter.
Ex.
"res1" is owned by "client.id" and given "update" scope
permission to user "usr" via policy/permission combo
$TOKEN is the access token for user "usr"
curl -X POST \
https://something/auth/realms/something/protocol/openid-connect/token \
-H "Authorization: Bearer $TOKEN" \
--data "grant_type=urn:ietf:params:oauth:grant-type:uma-ticket" \
--data "audience=client.id" \
--data "permission=res1.id#scope" //correct RTP with "update" for
"res1"
OR
--data "permission=res1.name#scope" //correct RTP with "update" for
"res1"
When a resource is owned by a user, you can only get a RTP by providing the resource id
(uuid) in the "permission" parameter. Requesting by name returns an
"Resource with id [res2.name] does not exist."
Ex.
"res2" is owned by "usr" and has an "update" scope
$TOKEN is the access token for user "usr"
curl -X POST \
https://something/auth/realms/something/protocol/openid-connect/token \
-H "Authorization: Bearer $TOKEN" \
--data "grant_type=urn:ietf:params:oauth:grant-type:uma-ticket" \
--data "audience=client.id" \
--data "permission=res2.id#scope" //correct RTP with "update" for
"res1"
OR
--data "permission=res2.name#scope" //"Resource with id [res2.name] does
not exist."
The interesting thing is that If you request a RTP without specific "permission"
property, keycloak returns the correct RTP with "update" for both res1 and res2
as it should.
Our tests also shown that this behavior does not rely on the "user managed"
property but only the "owner" property
Is this supposed to happen?
If yes, why?
If no, which one of the two is the buggy behavior? The behavior for the user owned or the
client owned resource?
The main reason for this email is that the fact that you can obtain RTP based on resource
name is immensely helpful for us since the other clients (other than the resource
provider) cannot get the resource id from keycloak but they do know what they are looking
for (the resource name). Not being able to get RTP based on resource name for user owned
resources, forces us to use a generic RTP for all resources every time which could become
a burden if a user can access a very large number of resources.
Best regards,
Kyriakos Stefanidis