Policy Triggers Evaluation
This design note was written mainly because of MID-8269 and its siblings (MID-8243, MID-8251, maybe others).
The issue is that rules dealing with segregation of duties (the
exclusion policy constraint) are explicitly allowed to be defined "uni-directionally", i.e. on the one of the two excluded roles.
The definition should then be applied bidirectionally.
Currently, this is the case only for
enforcement policy action.
The role approval and pruning still require the exclusion to be defined in both directions.
What Needs to Be Done
Technically, we need to browse through all exclusion triggers and somehow "attach" them to both evaluated assignments, not only to the one to whom they are physically assigned/induced.
However, current API does not provide a clean mechanism for this, as it only deals with the policy rules, not individual triggers. See the source code.
Therefore, this process needs to be done at each place where triggered policy rules are "consumed", i.e. in
Other places where it would be potentially needed in:
Obviously, this is not good. We do not want to repeat the same (or very similar) code needlessly.
Related Problem 1
There is another problem: When looking for exclusion constraints, we need to crawl through the rules and their (potentially composite) triggers. This needs to be done e.g. in request access UI code and in pruning code. Although the code itself is simple, it is not well maintainable in the long run.
Related Problem 2
Generally, when we have a policy rule, and we want to act somehow upon its triggers, it is hard to distinguish which triggers are "substantial" and which ones are "auxiliary".
The former represent the core of the rule, like the exclusion constraint or "assignment being added" one.
The latter may represent e.g. conditions under which the rule is to be applied.
The current specification language does not distinguish between the two.
We only have
presentation/hidden flags that drive the presentation of the condition(s) to the user.
(Partial) Solution Proposal
We should improve the API (e.g.
EvaluatedAssignment) to be able to provide information like this:
Get all exclusion triggers (also the ones triggered from the opposite side), with the ability to easily detect e.g. the requested policy actions for each of them. This would be useful for request access UI and the pruning.
Get all policy rules triggered from the opposite side of the exclusion. This is useful for approvals.
The problem 2 is still to be solved.