[xacml] RE: [CR/Proposal] Selector usage in match and not introducingnew concepts in function interface

[Daniel Engovatov <dengovatov@crosslogix.com>]

Issue with proper usage of Selector and Designator elements  as arguments
to the function used in target Match can be clarified with followings
statement:

"When <Selector> or <*Desinator> elements are used as an argument to a
function, that requires a single value of some type as the argument, and
the result of selection can not be converted to this type: either due to
incorrect syntax in the presented context, or due to multiple occurences
of value in context, value of the argument is INDETERMINATE"

For those concerned with compile time type safety: when XACML document is
processed, the type of the argument is inferred from the function
definition. In the similar way, when <selector> or <designator> is used
as an argument to a function requiring single value as an argument, it
shall be treated as if *-first-and-only() function is applied to it. So
some "preprocessor" transforms for a function foo(typeX), when used as
foo(<selector>) into foo(typeX-first-and-only(<selector>)). Operational
behavior is the same.  This application is inferred from usage in the same
way as DataType is inferred.

Systems, used with runtime type safety checks, for both correct type and
cardinality, may skip this step when interpreting XACML policy.

Such approach preserves simple extensibility interface for functions, 
especially usefult with systems that provide extensibility point for custom
functions not restricted to a single implementation language, or with common
component interfaces.

No schema modification or introduction of new concepts in needed.


daniel;