[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Dynamic Attributes versus Existing Attributes
All, I've been considering the issue of how to handle dynamic attributes generated by the Dynamic Attribute Authority that are attributes already in the request context (see Section 3.4 of https://www.oasis-open.org/apps/org/workgroup/xacml/download.php/68861/xacml-3.0-dyn-attr-v1.0-wd03.docx ). The three main options are "merge", "dynamic attribute overrides" and "existing attribute overrides". There are straightforward ways to get the effect of "merge" or "existing attribute overrides", whatever the default behaviour, by judicious use of the issuer field and the existing DAA obligations, but not in the case of "dynamic attribute overrides". This would suggest that "dynamic attribute overrides" should be the default, though it feels a bit like a sledgehammer and isn't the path of least resistance. XACML allows the same attribute to appear multiple times in the request, with one or more possibly-duplicated attribute values in each instance, so tacking on a few more generated by the DAA is trivially easy and is likely to involve the least disruption to existing context handler implementations. This is the "merge" option. The current DAA draft merges while avoiding adding duplicate values, but it would be simpler to not worry about duplicates since they rarely matter and there are a few ways to avoid them if required. If "merge" is the default then I would propose adding a new obligation so as to get the effect of "dynamic attribute overrides" when desired. Call it the ignore-initial-values obligation. It would have the same components as the exclude-all-values obligation and its effect would be to cause the context handler the exclude any original values of the nominated attribute from the request context. If I don't hear any arguments to the contrary, in the next DAA draft I will keep "merge" as the nominal behaviour, except in the presence of an ignore-initial-values obligation, and drop the need to check for duplicates. I'll add a non-normative section on ways to get the different behaviours. Regards, Steven
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]