OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

# xacml message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]

Subject: Re: "else" is what ? Was:Re: [xacml] Generalizing on-permit-apply-second

• From: Erik Rissanen <erik@axiomatics.com>
• To: Steven Legg <steven.legg@viewds.com>
• Date: Thu, 23 May 2013 09:14:46 +0200

```Hi Steven,

```
I haven't thought about the details yet. Superficially it's deny or N/A which would cause the third policy to be evaluated, but I haven't thought about all the corner cases and their implications.
```
```
I would prefer that the whole policy set is indeterminate if it's ambiguous which child gets evaluated, rather than combining both of them. The point is to be able to select on and only one of the children using a condition. Combining them both in case of errors contradicts the original requirement.
```
```
```
Best regards,
Erik

On 2013-05-23 02:04, Steven Legg wrote:
```
```
Hi Erik,

In the current draft the second child is evaluated if the first child
```
evaluates to Permit, Indeterminate{P} or Indeterminate{DP}. In the extension
```to if-then-else, what decision(s) from the first child would cause the
third child to be evaluated ?

Equivalence to your "today" example would mean that we evaluate both the
```
second and third children and combine the results if the first child evaluates
```to Indeterminate{P} or Indeterminate{DP}. We evaluate the third child and
not the second child when the first child evaluates to NotApplicable.

Regards,
Steven

On 23/05/2013 12:21 AM, Erik Rissanen wrote:
```
```Hi Bill,

```
You did not understand me. What I am saying is that because there is no combining algorithm like the one I am proposing, one has to use a redundant condition to prevent the second policy to be evaluated. Today you
```have to do this:

PolicySet [first-applicable]
PolicySet [on-permit-apply-second]
Policy
Rule
Condition foo
PolicySet
-> some content A (the "then" branch)
PolicySet [on-permit-apply-second]
Policy
Rule
Condition not(foo)
PolicySet
-> some content B (the "else" branch)

```
Note the line saying "Condition not(foo)". It is there to prevent policy set B to be evaluated in the case the condition foo is true, but policy set A is NotApplicable. I want instead to write:
```
PolicySet [on-permit-apply-second]
Policy
Rule
Condition foo
PolicySet
-> some content A (the "then" branch)
PolicySet
-> some content B (the "else" branch)

```
Regarding use cases, I think it's a valid requirement to be able to use an arbitrary condition to choose between two arbitrary policies. This shows up often and there is no convenient way to do it today.
```
Best regards,
Erik

On 2013-05-22 15:54, Bill Parducci wrote:
```
Not to be pedantic, but that is an implementation not a use case. I understand that you cannot divulge proprietary information so I will not press further. That said, the concept of using a condition from one policy set to negate ("protect") another seems wrong to me. Perhaps I am alone in this perception. I would be interested in knowing how this is arbitrated when numerous policy sets each negate the other.
```
Thanks

b

```
On May 21, 2013, at 11:45 PM, Erik Rissanen <erik@axiomatics.com> wrote:
```
```
```Hi Bill,

```
Yes, I already did: "if condition foo is true then use policy set A for making the decision, otherwise
```use policy set B"

```
I have many customers who have requirements in this form: some set of conditions which depend on the subject or resource or a combination of both, which determine which policy set applies. It could be about for instance different government regulations which apply in different cases depending on the
```subject/resource.

```
I cannot be more specific for two reasons, 1) confidentiality of customer data, 2) the specific use cases really are very diverse, but they all have the same form, that is, to choose among different policies
```based on a set of conditions.

```
The reason that it works poorly in XACML today is that all combining algorithms (except only-one-applicable) "backtrack" out from policy set A if it turns out to be N/A. However, in these cases the customer wants to make sure that policy B is never applied if the correct policy is policy A.
```
```
The problem can be solved by "protecting" policy B with the negation of condition "foo", but that leads to redundancy in policies, which puts an overhead on either authors or tools to manage it.
```
Best regards,
Erik

On 2013-05-21 18:40, Bill Parducci wrote:
```
```Can you provide an example of your use case?

thanks

b

```
On May 21, 2013, at 9:16 AM, Erik Rissanen <erik@axiomatics.com> wrote:
```
```
```Hi Bill,

```
Ok, I understand what you mean, but that is not the use case I am trying to solve. What you say is a vastly restricted subset of the general requirement "if condition foo is true then use policy set A for
```making the decision, otherwise use policy set B".

```
It could be something like "if the subject is an internal employee or an external person with temporary clearance for internal access, use the internal need-to-know policy (which is very complex), otherwise use the policy for external information sharing regulations compliance (which is also very complex)"
```
```
It's only rarely that you can qualify the resource into a partition of classes like in your example.
```
Best regards,
Erik

On 2013-05-21 18:00, Bill Parducci wrote:
```
Before we dive down into the syntactical weeds I suggest we look at this from a simpler perspective. I
```see it something like this:

```
Bob requests access to Resource Foo. Foo exists in 3 flavors: FooBronze, FooSilver, FooGold. The PEP generates a request to the PDP for Bob to access Foo. (Note that the request has no references to the instances of Foo.) The PDP performs whatever level of complex processing that the language allows. Assume the decision is Permit and there is an Obligation tied to some attribute of Bob (or time of day, network speed, concurrency, behavior, etc.) This Obligation states OnPermit::Foo:FooSilver. The answer the PEP receives is Permit;Foo:FooSilver. It then grants Bob access to Resource FooSilver.
```
```
This requires that the PEP has the ability to process an abstract Resource, which implies a non-linear access control mechanism (routing, redirection, etc. vs. "locked gate"), which may or may not apply to the system you are describing. It is a common mechanism for solving such problems which is why I am throwing it out there. A "whitebox" understanding of the decision process with the PDP as currently defined this aspect is not new and this solution doesn't require the PDP to have any understanding
```whatsoever of the decision making process itself.

Does that make sense?

b

```
On May 21, 2013, at 8:25 AM, Erik Rissanen <erik@axiomatics.com> wrote:
```
```
```Hi again Bill,

```
To add more thing. The reason I don't understand how obligations relate to this is that obligations are processed on the PEP side. I want the "if condition foo is true then use policy set A for making the decision, otherwise use policy set B" to happen on the PDP. So I cannot see that there is any way an obligation can impact this, unless you intend somehow to defer policy evaluation to the PEP side.
```
```
Also keep in mind in your example that "foo", "policy set A" and "policy set B" can be arbitrarily complex. I just used the role example to keep the example size small.
```
Best regards,
Erik

On 2013-05-21 17:17, Erik Rissanen wrote:
```
I don't understand how an obligation can solve what I perceive as a missing feature in the policy
```evaluation logic.

```
Basically, what I see is that customers have requirements that say "if condition foo is true then use policy set A for making the decision, otherwise use policy set B".
```
```
Could you elaborate in more detail in pseudo code XACML how your solution works? I do not understand what you wrote. Could you specify it with policy set structure and combining algorithms?
```
Erik

On 2013-05-21 17:08, Bill Parducci wrote:
```
I suggested a more general solution in my first note that I don't think will require a core schema change: Create an abstract Action that is instantiated by an Obligation. That should address your issue as well provide a more robust solution than if/then, if/then, if/then, if/then, if/then... without the need to embed ordered Boolean logic chains within rule sets. I'd be happy to discuss the idea with your in more detail to see if it might solve this issue as described.
```
```
You are free to create and propose whatever you would like. However, what Axiomatics chooses to
```build outside of the spec is not really relevant to this list.

b

```
On May 21, 2013, at 12:19 AM, Erik Rissanen <erik@axiomatics.com> wrote:
```
```
Sure, I can see how a more general switch type of expression could be useful, but that does not mean that an if-then-else is not useful in itself. I can see a big need for it from customer deployments. And I don't want to do core schema changes, so a combining algorithm extension will
```work nicely.

```
If the TC is against it in the combining algorithms profile, I'll just do a custom Axiomatics
```extension for this. It's important enough for me.

```
What would you suggest as the more general solution? The on-permit-apply-second could be generalized even further so it can take a number of pairs of policies as children, like this:
```
PolicySet on-permit-apply-second
Policy - Rule - Condition A
PolicySet -> used if A was true
Policy - Rule - Condition B
PolicySet -> used if B was true
Policy - Rule - Condition C
PolicySet -> used if C was true
... etc

```
Optionally there can be a final single child which will be used as the "else" case if nothing else matched. I think this should solve the more general case you are pointing to.
```
Best regards,
Erik

On 2013-05-20 18:14, Bill Parducci wrote:
```
Your are proposing if/then but I still maintain that the use case is unnecessarily limiting. A simple A|B Boolean construct being added doesn't address what I believe is the underlying A|B|C
```(D|E|F...) issue to be solved.

```
It seems that Policy authors would only be affected in either solution via the implementation of the client facing editor. Since there isn't any branching logic in the spec now any change would
```by definition be new.

```
I believe that if we are going to go through the effort of getting something like this standardized we should consider taking the more generalized approach.
```

b

```
On May 19, 2013, at 11:51 PM, Erik Rissanen <erik@axiomatics.com> wrote:
```
```
```Hi,

```
I am asking for an explicit "if-then-else" into XACML. "if-then-else" is often how people formulate their access policy requirements and it's an overhead for them to have to re-factor
```the requirement into some other form of expression.

Best regards,
Erik

On 2013-05-17 17:03, Bill Parducci wrote:
```
I am not suggesting that the outcomes be expanded. I am saying that it appears your are effectively asking for a polymorphic Resource that is bound to a Subject. The Action is unchanged re: access control. An if/then mechanism is not the way I would handle it. What if you want to differentiate between senior employees or admins vs. general employees or others?
```The solution becomes unwieldy.

Subject: user employee
```
PolicySet 1 -> Permit/Deny (on abstract Resource); Obligation 1 (Resource instance 1)
```
Subject: user manager
```
PolicySet 1 -> Permit/Deny (on abstract Resource); Obligation 2 (Resource instance 2)
```
```
PolicySet 1 -> Permit/Deny (on abstract Resource); Obligation 3 (Resource instance 3)
```
[...]

No boolean machinery is needed.

b

```
On May 17, 2013, at 7:30 AM, Erik Rissanen <erik@axiomatics.com> wrote:
```
```
```Hi Bill,

```
I don't see any relation to obligations at all. I'm simply thinking of things like:
```
if "user is employee" then
PolicySet 1
-> policies about employees go here
else
PolicySet 2
-> policies about all other kinds of users go here

```
This is for making the decision regarding Permit/Deny. It's not about expanding the possible
```outcomes to a larger set than Permit/Deny/NA/Indet.

Best regards,
Erik

On 2013-05-17 15:42, Bill Parducci wrote:
```
This is a subset of the QoS use case I have advocated for in the past. I personally believe that this is best handled via an Obligation: the Permit is granted to an abstract Resource and the Obligation defines the Resource Instance. This keeps the decision machinery simple, while allowing for trinary, quaternary... decisions rather than being limited to an if/then
```construct.

Permit: Resource {foo}
Obligation: foo=>foo1

b

```
On May 17, 2013, at 12:54 AM, Erik Rissanen <erik@axiomatics.com> wrote:
```
```
```All,

```
I have a suggestion to generalize the on-permit-apply-second algorithm.
```
```
The problem I see is that it's not unusual for customers to choose among policies on a "if then else"-basis. In other words, if a condition is true, choose one policy, otherwise use
```another policy.

```
The problem is that there is no combining algorithm which does that currently. Yes, we do have targets and conditions and algorithms such as first-applicable, but there is a crucial difference between "if then else" and these combining algorithms. Consider this case:
```
PolicySet [first-applicable]
Policy target foo
-> some content A
Policy
-> some content B

```
Naively this is an attempt to write an if-then-else type of policy set. The condition is "foo". If foo is matching, then "some content A" is evaluated. However, the problem is that if "some content A" itself returns N/A, then the first-applicable combining algorithm will evaluate "some content B". So, even if the target "foo" is matching, it can turn out that "some content B" is used. This is not what "if then else" means.
```
```
There is the only-one-applicable which is closer to if-then-else, but it works only on
```targets, not on conditions.

```
The way to model a condition based if-then-else in XACML today is to use the on-permit-apply-second and invert the condition in the "else" branch. Like this:
```

PolicySet [first-applicable]
PolicySet [on-permit-apply-second]
Policy
Rule
Condition foo
PolicySet
-> some content A (the "then" branch)
PolicySet [on-permit-apply-second]
Policy
Rule
Condition not(foo)
PolicySet
-> some content B (the "else" branch)

```
As you see, it takes a whole lot of XACML elements to do a simple if-then-else, which is a
```very common use case.

```
My proposal is to generalize the on-permit-apply-second algorithm so it can take a third policy. If there are only two children, then it behaves like in the current spec. However, if there is a third child, then the third child is applied if the first policy does not
```return permit. Then you can do the following:

PolicySet [on-permit-apply-second]
Policy
Rule
Condition foo
PolicySet
-> some content A (the "then" branch)
PolicySet
-> some content B (the "else" branch)

```
This is more compact and efficient and does not contain redundant expressions.
```
```
Let me know if you think this is a bad idea. Otherwise, I can work out the details into a
```new working draft of the profile.

Best regards,
Erik

```
--------------------------------------------------------------------- To unsubscribe from this mail list, you must leave the OASIS TC that generates this mail.
```Follow this link to all your TCs in OASIS at:
```
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
```

---------------------------------------------------------------------
```
To unsubscribe from this mail list, you must leave the OASIS TC that generates this mail. Follow this link
```to all your TCs in OASIS at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
```
```
```
```
```

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]