[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [xacml] Comments on XACML Privacy Policy v1.0 profile
My opinions on some of the points raised in this thread. Undesirability of black lists. In the abstract, the argument goes back to the earliest days of XACML. It has appeared in many forms. Should negative policies be allowed? Should combining algorithms like default allow be prohibited? Should policy creation tools prohibit “impossible” or “illogical” policies. I believe you can find debates between me and Bill on subjects of this sort dating back to 2001. That is before XACML 1.0 existed. The approach the TC has always taken is to NOT limit the power of the language, just because we can’t see a use for a certain construct doesn’t mean there isn’t a legitimate one. In addition there has generally been a sense that although there might be a small number of truly useless policies, (“A” and NOT “A”) it is not worth the trouble to try to figure out what they are and require everyone to prohibit them. Of course it is perfectly legal for editing tools to detect questionable constructs and issue warnings or suggestions. Consistent with this, I think it is ok to mention the use of blacklists while warning of theuir dangers. I question the use of a policy as normative in any Profile. In practice, the policies even if written with the same intent there are bound to be small deployment–specific differences. For example, what If I add an Issuer? What if I set must_be_present to “true”? Without a detailed set of criteria, how can I determine if a given policy is compliant? I would prefer to say” You must check such and such her is a sample of a policy that does the right thing. As far as exhibiting Privacy policies in general, I think it would be a lot more useful to create a library of rules and policies which are useful for checking Privacy. We could for example start with the ones from the XSPA interops and then add ones to cover the cases Mohammad has suggested. In my view the key thing in the policy that needs to be normative is the syntax and semantics of the attributes and any datatypes or functions defined in the Profile. In any event, I would like to get consensus on whether changes are required and who is going to propose them. Hal From: Erik Rissanen [mailto:erik@axiomatics.com] Hi Hohammad, On 2014-06-05 23:26, Mohammad Jafari wrote:
|
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]