Difference between revisions of "Requirements Model Part Three"
|Line 147:||Line 147:|
You will need a Bugzilla account to add comments.
You will need a Bugzilla account to add comments.
Revision as of 14:22, 21 February 2009
- 1 Introduction
- 2 The high level view of the model
- 3 KnownType/Discriminator pattern
- 4 Users and Roles
- 5 System and related elements
- 6 Requirements elements
- 7 Comments
Mindful of all the precious comments and suggestions we have been receiving in the last few weeks, we have now put together a brand new requirements model, which we affectionately call model 2.0 and which is described in the sections below. We hope you all like it (we do!... :-), but we vividly encourage all ideas and criticisms.
A PDF version of the model description is available. So, if you prefer to review the model from the comfort of your favourite armchair, just click here.
The actual model, which was captured with Soyatec's eUML tool and is provided in files in UCD format, may be accessed from the SVN repository. Soyatec offers a free version of the tool, which is the one we are using. The repository is located at <https://dev.eclipse.org/svnroot/technology/org.eclipse.ormf> and the project that contains the UCD models is <trunk/analysis/org.eclipse.ormf.analysis.model2>, with the actual UCD files residing in the "models" folder.
As usual, please feel free to add discussion points below the model description.
The model builds on top of the conversations that have taken place in page Requirements Model and Requirements Model Part Two of this wiki. All ye readers may want to peruse those pages to get a background on the reasoning behind this new model.
The high level view of the model
At this stage the model simply reflects the main elements and their mutual relationships.
The figure below shows major players and their most fundamental associations.
These elements are described in the rest of this section.
At the highest level of abstraction we have NamedElement (to the right of the diagram above), which represents the most generic element in the model. All model elements are minimally characterised by a name.
The NamedElement is specialised by the ReferenceableElement, which, as its name implies, is an element that may be referenced by any other element in the model.
We then have the ManagedComponent element, which is unarguably at the core of the model. A ManagedComponent is the smallest ReferenceableElement that is managed directly by the framework. As such, ORMF guarantees to provide all basic CRUD operations on this element, as well as the direct persistification into the repository of the system. This is to be contrasted with other NamedElements, such as a Relationship or an internal constitutive element of a ManagedComponent, which are only created, deleted or persisted as a result of the creation, deletion of persistification of the ManagedComponents to which they contribute.
Furthermore a ManagedComponent is the only element in the model which has the concept of, and an association to, one or more Revision.
ManagedComponents are associated to one or more User playing a specific Role, with the single required Role being that of the Owner. More details on this relationships will be provided in section Users and Roles below.
Given its central position in the model, ManagedComponent is the abstract parent of a large part of the elements that comprise an ORMF based system.
System is a ManagedComponent that represents the entire system that is being created and managed with ORMF. From a containment perspective, the System is the root container of all elements in a requirements model. See section System and related elements below for details on the associations of System with the other elements.
User is a ManagedComponent that encapsulates the generic user of an ORMF based system. All issues of authentication and authorisation that are typically connected to a user are handled by ORMF. A User has one or more Roles to play in association with any ManagedComponent. More details on this relationships will be provided in section #Users and Roles below. User elements are contained within the System element. See section System and related elements below for details on the associations of User with the other elements.
Project is a ManagedComponent that contains all ManagedComponent elements created to describe the requirements model of a specific software project. Project elements are contained in the System element and may in turn contain any number of Folder elements. See section System and related elements below for details on the associations of Project with the other elements.
Folder is a ManagedComponent which acts as a container of other Folder elements or ManagedComponents. If not contained by another Folder element, a Folder sits within a Project. See section System and related elements below for details on the associations of Folder with the other elements.
RequirementArtifact is a ManagedComponent which is the common parent to all elements that represent some form of requirement.
RequirementArtifact has two specialisations, which will be described in more detail in section Requirements elements below. Section Requirements elements describes also how we see customised RequirementsArtifact elements being added to the system by any adopting organisation.
A RequirementDocument is a ManagedComponent which represents a textual artefact that acts as a “wrapper” around specific subsets of ManagedComponents.
A good example of this type of element is the Software Requirements Specification document, which is typically a high level survey of all functional, non functional and use case related requirements for a software project or for a specific component of the project. The analyst typically organises the survey in a particular way, that needs to be reproducible through the natural evolution of the requirements and that needs to contain sections of descriptive text that acts as a cohesive agent for all requirement artefacts proper. Examples of such sections are introduction, overview of the problem domain etc. This text is what the model refers to as Augmentation Text, which is a separate model element, not shown in the diagram above; any Requirement Document makes use of an arbitrary number of Augmentation Text elements.
ExternalArtifact is a ManagedComponent which embodies any external resource (such as an external document, an image or a diagram) that is referred to by (or included in) one or more of the ORMF ManagedComponents. Once the User makes ORMF aware of an ExternalArtifact, it ensures that the artifact exists and can be found when needed.
KnownType and Discriminator
These elements are both ManagedComponent elements. Together they form the basis of the ubiquitous pattern that facilitates extensions of the ORMF framework. Please see section KnownType/Discriminator pattern for a description of these two elements and of the related usage pattern.
An Elucidator is a ManagedComponent that, although not a requirement artefact in its own right, acts as a means to clarify or enrich any NamedElement, especially the various Requirement types.
Elucidators are elements that may be utilised across a number of different NamedElements and that permeate the entire requirements model. Examples of Requirements Elucidators are Issues, Glossary Terms and Notes.
The type of a specific Elucidator is defined via the KnownType/Discriminator pattern with KnownType = Elucidator, which is described generically in section KnownType/Discriminator pattern. This ensure easy extensibility of the Elucidator element.
A diagrammatic representation of these two elements and their associations may be found in the top diagram in this section above.
Relationship is a NamedElement. It represents the association between any NamedElement and any ReferenceableElement, where the NamedElement is the source of the association and the ReferenceableElement is the target.
NamedElement and ReferenceableElement objects may have any number of Relationships defined.
Relationship also carries the concept of an association’s stereotype through its usage of a specific Discriminator object which defines the desired value for a KnownType = Stereotype. The KnownType/Discriminator pattern is described generically in section KnownType/Discriminator pattern. This ensures easy extensibility of stereotypes.
Revision is a NamedElement which contains all the required information related to revision management and control of any ManagedComponent. A ManagedComponent may be associated with any number of revisions.
KnownType is a ManagedComponent element which identifies a type’s name, which is indicative of the particular usage being made of the pattern. Examples of KnownType names are “Stereotype”, “Elucidator”, “Requirement”, “Narrative”, “Role” and “Priority”.
Discriminator is a ManagedComponent element which identifies a possible value for a specific KnownType. When a particular KnownType element may accept more than one value, there will be a Discriminator object for each required value. For example, for the KnownType “Role”, there will be, out of the box, at least four Discriminator objects, with values of “Stakeholder”, “Owner”, “Administrator” and “Project Administrator” respectively. If an adopter requires to identify a new role to be associated with a User, the only things that is needed is to contribute a new Discriminator for KnownType = “Role”, with the value of the Discriminator being the desired value. All of the semantics that is necessary to handle the concept of a role is automatically provided by ORMF and therefore ORMF’s core does not need to be modified.
The same mechanism can be utilised to extend the system to accept a customised requirement, provided it fits into either the Requirement or the Narrative categories.
Besides adding Discriminators with custom values to existing KnownType elements, adopters may also create their own KnownType and associate one or more Discriminator elements to them. This in itself will be sufficient in most customisation circumstances, as the Discriminator element offers all the required behaviour to accommodate for searching, reporting and filtering based upon specific Discriminator values. We see this as a very powerful tagging mechanism, that will enable adopters to provide specialised visual views into a requirements model or specialised reports or documentation.
Clear examples of popular attributes that can easily be handled simply by the KnownType/Discriminator elements are various management related attributes, such as a requirement’s status, priority or complexity. So, for example, a KnownType with name “Priority” may be created, along with three Discriminator objects with values “low”, “medium” and “high” respectively. Any ManagedComponent may then simply be tagged with the Priority KnownType and its suitable corresponding Discriminator. All desired selecting, filtering and reporting based upon the Priority KnownType will then be immediately available and easy to activate.
The management related KnownType mentioned above will be included in the framework, but many other may be envisaged based upon a development team’s habits or a Project Manager’s categorisation preferences.
In those cases in which a more specialised and unique behaviour than the one offered by the plain Discriminator is required, custom elements embodying this behaviour can be created. Examples of such elements in the ORMF model are the Role element and the Relationship element.
Users and Roles
The figure below shows a diagrammatic representation of the concept of Role in the model.
The Role element embodies all of the semantics that is expected to be associated with the role concept.
As the diagram shows, a Role is only valid in the context of the ManagedComponent with which the User is interacting. Consequently, a User may possess many different Role objects for different ManagedComponents.
Each instance of the Role element must be associated with a Discriminator for KnownType = Role that defines the specific role type for that instance of Role.
ORMF comes with a number of pre-defined role types, but the list may be easily extended using the pattern described in section KnownType/Discriminator pattern.
The figure below shows the containment relationships between System, User, Project, Folder and ManagedComponent elements.
As already mentioned, the System is at the root of the containment hierarchy. It contains any number of User elements and any number of Project elements. Each Project, in turn, contains a number of Folder elements, each of which may contain either more Folder elements or other ManagedComponent objects.
A Folder always knows about its containing Project, regardless if the Project element is its direct parent or not.
The figure below focusses on those artefacts that directly represent some form of Requirement.
The basic element embodying a requirement of some type is the RequirementArtifact. This is the most generic requirement related element and it is specialised by two types, namely Requirement and RequirementNarrative.
The Requirement element represents such artifacts as a Functional Requirement, a Constraint or a Quality Attribute.
The RequirementNarrative elements represents text centric artifacts such as a Use Case, a User Story or a Scenario.
All of the behaviour that is common to requirement artifacts in the categories of Requirement or RequirementNarrative respectively is specified in these two elements.
Requirement and RequirementNarrative take advantage of the KnownType/Discriminator pattern by indicating their specific type via a particular Discriminator associated with KnownType = Requirement or KnownType = Narrative respectively (see section KnownType/Discriminator pattern for details on this pattern).
Discriminators for KnownType = Requirement that are known to ORMF out of the box are inclusive of Business, Constraint, Functional, Performance, Quality of Service and Stakeholder requirement.
Discriminators for KnownType = Narrative that are known to ORMF out of the box are inclusive of Scenario, Stakeholder Need, Use Case and User Story.
Furthermore a Requirement element is characterised by associations with other Requirement elements which can be summarised as defining Requirements that are derived from a source Requirement and/or Requirements that provide greater detail to an abstracted Requirement.
Finally we have the concept of a CompositeRequirement, which is a specialisation of the Requirement element that acts as a container for other Requirement elements.
Please enter comments as discussion points below by selecting the edit icon to the right. It would be helpful if you could title your discussion point.
You will need a Bugzilla account to add comments.
Questions on the model
After trying to make sense of the model all by myself for some time, I think it is worthwhile to ask some questions.
Overall, I have difficulties understanding the model, since the explanations given do not explain the intention or uses of these elements, and I do not find the model self-explanatory or intuitive. My questions may very well be completely off due to my lack of understanding.
Q: Use of System element
Can you please explain the use of this element? Why would I need it? Does it represent a concrete (technical) system, an endeavour? What else? I also struggle with the fact that the User element is contained by the System element. Does it mean that one user must be present multiple times, once for every system she is asociated with?
Q: How to differentiate multiple relationships?
If each Relationship is only associated with one Discriminator (which, as I read from the documentation, represents a concrete value of some kind), how can relationships defined by more than one criteria be handled (for instance "priority = high" and "status = open")?
Q: Role assignment
Why can a Role only be associated with one single User? Can't there be, for instance, multiple authors of a document? Or multiple reviewers? What is the use of the singular Discriminator for the Role element?
Q: Use of Discriminator
Why can only ReferenceableElements have a Discriminator? Are Discriminators only used in relationships? And only on the target's side? I must admit, the whole use of these needs a lot more explanation to make sense to me. You describe the "KnownType/Discriminator" pattern using examples such as "Requirement", or "Role", which are primary model elements themselves - can you explain, why one would want to have such a KnownType instance? It actually confused me more than it explained. Further down, I understand your explanation as those being "simple" tagged attributes, or "properties" as they are often called in Java. I am not sure if my understanding is correct here, but if so, should't then the model deal with the various types of properties that can exist, such as numeric properties, enumerated values, strings, etc.? Maybe a concrete example of how you think they should be used would help (including an object diagram to show the runtime instances).
I would really appreciate, if you could possibly add more information on how you think this model should be used. I am sure things will become much clearer once you describe what each element is for and how it is to be applied in ORMF.
--Ormf.ib-ponikwar.de 18:22, 21 February 2009 (UTC)