UTP 2 defines for each element that may produce a verdict (i.e. test set, test case and all procedural elements), default arbitration specifications that prescribe how to calculate that verdict. The default arbitration specifications are optimized for functional testing, though. In case of testing non-functional aspects such as fault tolerance, performance or security, the rules to determine functional verdicts do usually not suffice anymore. Therfore, UTP 2 enables the definiton of proprietary arbitration specification (mainfested as stereotype <<ArbitrationSpecification>>) to grant the hightest flexibility possible to the user.

However, the current mechanism to associate a test set, test case or procedural element with a dedicated arbitration specification is restricted to only one arbitration specification. Even though this would be sufficient in many situations, the problem with the current binding mechanism is that the test set, test case or procedural element is literally changed by the binding. This may lead to situations where elements are copied only for the sake of associating a different arbitration specification to a that element. This is a situation that should be avoided under all circumstances.

We suggest adding a loosely coupled arbitration specification binding mechanism that keeps the elements and the arbitration specifications that ought to be bound to that element separated from each other. This would bring the benefit that test sets, test cases and procedural elements stay clean of any arbitration specification. This would also foster reusability of these elements.

The loosely coupled mechanism to bind arbitration specifications to the test sets, test cases and procedural elements could be based on a new concept 'ArbitrationDirective' , similar to the 'TestDesignDirective' concept that already exists in the language and proved its applicability. By doing so, users could define and assign an arbitrary number of arbitration specifications to a test set, test case or procedural element without technically modifying any of these elements.

The last paragraph of Clause 8.6.2 Data Values is an unintended left over from the revised submission. In the revised submission, there was the attempt to precisely define the semantics of matching ValueSpecifications. This attempt ought to be entirely removed. Thus, this paragraph should be removed.
For the sake of clarity:

Both stimuli and expected responses yield data values for distinct signature elements. A signature element is defined
as instance of either a Parameter or Property (i.e., this specification introduces a virtual metaclass SignatureElement that is the joint superclass of Property and Parameter and has at least the following attributes: type : UML::Type,
lower : Integer, upper : UnlimitedNatural). Given by UML [UML25], a "... Type specifies a set of allowed values
known as the instances of the Type." This specification denotes this set in the context of a SignatureElement
expressed as type(se), with type(se) as SignatureElement.type, and use T as abbreviation for type(se).
We specify
with se instance of SignatureElement and lower(se) as SignatureElement.lower and denote it by SE type.
A ValueSpecification V as an argument for a SignatureElement is specified as
These basic definitions are further used for the specific ValueSpecification matching mechanism extensions
introduced by UTP.

There is an obvious misplaced change bar marker in the beginning of section "8.6.1 Data Specification". It needs to be removed.

In the current spec, the abstract syntax diagrams reside in dedictated sub-sections directly underneath the conceptual context section (e.g., 8.7.2. Test Logging) they belong to. Stereotype definitions of ths conceptual context are grouped in a dedicated sub-section 'stereotype definitions'.

In UTP 2.1 RTF turned out that this structure does not support maintainability and modifiability of the spec. Everytime an additional diagram is added to the conceptual context section (maybe due to enhacements, clean-ups, better seperation and illustration of complex diagrams), the entire stereotype definitions need to be renumbered, because the respective diagram sub-sections and the 'stereotype specification' sub-section reside on the same level.

If the abstract syntax diagrams would be sub-sections of a section called 'Abstract Syntax', there would be no need to renumber stereotypes.

Therefore, I propose to make the spec more maintainable and stable by incoporating the following structure for every concepual context section in the spec.

8.7. Test Evaluation
8.7.1 ArbitrationSpecifications
8.7.2 Test Logging
8.7.2.1 Abstract Syntax
8.7.2.2 Stereotype Specfications

"Figure 8.15 - Test-specific actions Overview" needs to be cleaned up. Stereotype shapes are overlapping.

The current document generation process used to produce the UTP 2 specs is limited with respect to the illustration of property/association end modifiers (i.e., unique, ordered, redefines, subsets, read-only, union). According to UML, those modifiers shall be illustrated after der name and type and multiplicity of a property. The following grammar rule (taken from UML 2.5, Clause 9.54) specifies the textual notation of properties:
<property> ::= [<visibility>] [‘/’] <name> [‘:’ <prop-type>] [‘[‘ <multiplicity-range> ‘]’] [‘=’ <default>] [‘

’]

In the UTP 2 document generation framework the illutration of modifiers was circumvented in a way that is is close to, but different to the notation prescribed by UML. Although, the aligned notation of UTP 2 is okay from a readers point of view, it appears that the modifiers have not been consistenly illustrated. See two examples from the spec:

In section 8.3.2.7.17 TestDesignDirective , TestDesignDirective modifies are displayed before the name of the propery:

: RoleConfiguration [*]

Even though both ways are still good enough to comprehend, from an editorial point of view, one way should be preferred to other.

This linebreak can be removed.