Configuration Management Scenarios

Sort of a ghost town here
Active specification development is now at Change and Configuration Management TC

Want to contribute?

History View Links to this page 2013 August 14 | 12:48 pm

Cross-domain Baseline

The primary use case for this workgroup is the creation and management of cross-tool, cross-domain configurations and baselines - that is, configurations and baselines composed of versions of resources from multiple applications that have might have multiple approaches to their own internal version and configuration management, and where those resources might be from more than one OSLC domain, or might not be defined by any current OSLC domain.

A user needs to be able to:

Create and update configurations
Identify the version of a given resource within the context of a configuration
Establish a baseline (snapshot) of a configuration
Describe the contents of a configuration in terms of sets of changes with respect to some other configuration

Global Configurations

In a cross-tool, cross-domain environment, there may be multiple configuration management tools or systems involved. For example, a site may use a mix of SCM systems such as Rational Team Concert, Subversion, Git, etc., and may use other systems such as the linked data configuration management service behind Rational Engineering Lifecycle Manager.

A user needs to be able to:

Define global configurations that assemble, contain, or otherwise reference configurations managed by more than one configuration management tool; the referenced configurations are called contributions
Edit a global configuration, adding or removing one or more contributions, or replacing a contribution with a different version or variant
Use these global configurations in all the other OSLC Configuration Management scenarios

Legacy Scenarios from Core Workgroup

Create a baseline

As a software developer, I need to record the exact state of my SCM configuration for building a library that will be used by other software components, including the right header files, the binary build artifacts for various platforms, the README or release notes, etc.

As a requirements manager, I need to record the exact state of my fighter jet requirements following a review by the Air Force. Since requirements may change, I need this snapshot to be available at any time later, so we can review what the Air Force has or has not yet approved. Links between my requirements must be included in the snapshot.

As a user of a system, I need to treat a thing (resource, metric, etc.) or collection of things as a related group, and I want the system to record and/or freeze a copy at some point in time, or (for resources that are versioned in some way by their provider) record the version of the resource at that time. The state of links between resources must also be captured.

Modify and delete baselines

As a release manager, I need to make a formal release of a software product. The previously created baseline for the tested final build of the product should be marked and preserved, adding appropriate version numbers and comments to the updated baseline. After that, I want to delete several previous test and published baselines for earlier intermediate states of this new product release, and the associated intermediate versions of build artifacts, in order to save space and speed up future queries.

As a baseline manager, I need to be able to modify the properties of a baseline, and link that baseline to other resources such as test results.

As a baseline manager, I need to be able to delete or mark obsolete old and unused baselines, so they do not affect users’ operations and queries.

Control the use of a baseline, component-based development

As a software developer, I need to be able to mark my new library baseline as available to the testing group, but I do not yet want the new library to be picked up by other software developers unless they explicitly choose to do so. Once the library has passed testing, we need to be able to mark the baseline as suitable for general use

As a user of a system, I need to announce and/or control the availability of a collection of resources to some audience (such as a testing group, or another development group using component assembly, etc.)

As the process owner, I need to be able to set rules for when new baselines or new versions of baselines are picked up by the users of my system

As a user of baselines, I need to be able to decide which baseline or version of a baseline to use as a sub-component, library, or similar dependency (possibly within limits and policies determined by the process owner)

High-ceremony requirements-driven development

This scenario illustrates the way in which an immutable requirement specification is used to drive downstream development work.

System engineer (RM) writes requirement specification, baselines it, gets it reviewed and signed off.
1. [Optional] Sytem engineer responds to review comments and updates the specification. A new baseline containing those updates is created, and re-submitted for review and approval.
2. [Optional] Unwanted baselines are deleted.
Quality engineer (QM) works to this requirement baseline, creating tests, and creates links from those tests to the requirements specification in the context of that baseline.
1. [Optional] QA provides feedback on the requirements (rework required; goto 1.1)
Quality engineer creates a overall baseline consisting of all the requirements and all the tests, and all the links between same.
Test necessity and sufficiency is assessed with respect to that requirements specification, in the context of this overall baseline.
The baseline is published as an auditable workpackage/deliverable for that project.

Incremental Development Using Baselines

This scenario is a generalization of the previous one, “High-ceremony requirements-driven development”.

As a developer, I want to establish a configuration (work space, project, etc.) using my own resources under development plus one or more baselines of referenced resources. After some development and testing, I want to construct a baseline that represents my current state - making a baseline from my own resources and including the referenced baselines or resources from those referenced baselines. This process may be repeated by me; furthermore, other developers can use the baseline I established as the start for their incremental development.

Compare baselines

As a test engineer, I want to be able to see the new change sets included in a baseline wrt the baseline I tested last week, and I want to be able to trace those change sets back to any associated defects or requirement changes.

As a creator or user of baselines, I need to be able to compare two baselines, or the current state of the system vs. a baseline, and get a report of the differences, tracking those differences back to their root causes across multiple SCM domains

Scenarios for Predecessors of Configurations

Usage scenario however:

We need to track changes between configurations.

Why was the new configuration assembled?
What was done since the predecessing configuration?
What problems have been solved from an old configuration to the current one?
What features have been added from an old configuration to the current one?
When is a configuration “good enough” for us to release (planned changes are implemented) or be built?
What is the “newest and best”configuration?

In my use cases there are two different change traces that need to be done, both the simple one, what direct configuration items are changed since the last configuration (compare to a simple file diff pointing out rows that are different), but also the more complex delta, a complete trace of changes in all versions that have been created of configuration items in the configuration (for each row that is different in the “file diff” we need to step through each version between these configuration items and find change data).

One could discuss that a configuration should not have a predecessor property since predecessor should be “calculated” through the predecessor of each individual configuration item, but couldn’t the configuration in itself also be a configuration item as well? From outside I would like to be able to view the configuration as a stand alone versioned item, no need for the viewer to know how many subitems are inside the item.

If I have a predecessor property on all items, I will be able to use more efficient algorithms and cache to calculate what changes were solved on a component consisting of an arbitrary level of subcomponents trivially by only needing to compare versions of this component instead of having to compare versions of all sub components. I don’t need to know “the whole world”.

Or does the problems above have more to do with change management?

PLM Systems engineering change scenario

This follows the PLM Systems engineering change scenario and extracts CM related aspects of it. The scenario outline includes:

Assign & communicate the change request (a1, a2, a3)
    * Assign change request context
    * Submit change request
    * Locate change request from notification

There appears to be a number of contexts that might be associated with the change request:

The context against which the issue is being reported. This is typically going to be some immutable configuration representing a product, component, or subsystem. For example, a defect might be reported against a component with release 1.0.
A context representing the target in which the change will be delivered. For example, the fix for the defect in the previous entry might target release 2.0.
A context representing the workspace in which the latest appropriate versions of resources are provided as a starting point for applying the change.

The first type of context is usually expected against a specific version of some configuration. This implies that CM providers must support URIs that always resolve to a specific versioned resource. We might use the term absolute resource version URI for this.

The third type of context is likely to represent a configuration whose versioned members are the latest according to some context. In the example, a configuration that has the latest delivered changes for release 2.0. This implies that CM providers might support a URI that represents the pair of a concept URI plus a context. We might use the term composite resource version URI. Is there a better term? Consumers of this data should not assume anything about semantics. Such a URI might literally composed of these two parts (concept+context), or might be simply an opaque URI. All that is required is that the underlying CM provider should be able to resolve that URI to reference the appropriate version of the concept resource.

The context might represent a number of tool or domain dependent forms. For a PLM tool, it is likely to include some notion of effectivity. For an SCM tool, it might include a release, or stream, or workspace.

Apply request context to establish impacted requirements & implementation (a4, a5, a6)
    * Locate requirements in change request context
    * Create new revision of requirements context and reserve for editing
    * Open new revision of context

Steps a5 and a6 above imply a capability to perform a check out to create such a new version, and be able to reference that new version in other resources, such as a change request. This raises some interesting questions:

Is an OSLC CM provider required to support initiating the check out to create the new version? For example, a POST to the current immutable version might create a new version. Or would it be sufficient in phase 1 to support an appropriate link type but require that the user use the versioning tool to initiate the check out?
Should a CM provider support a delegated UI to allow a link to/from another resource to be created and stored on the configuration item?
What about versioning tools for which check out does not create a new version but rather creates some kind of lock? Such tools might only create a new resource on check in.