The Eclipse Dataspace Connector will employ SemVer for versioning and distinguish between the following releases as defined by the Eclipse Handbook:
- Major releases, which introduce API changes,
- minor releases, which add new functionality, but are API compatible with previous versions, and
- service releases, which include bug fixes only and add no significant new functionality.
Between releases, snapshot versions reflecting the current state of modules can be packaged to distribution artifacts on a regular basis. Snapshots, however, do not actually represent released versions.
The concept of API compatibility is defined in terms of binary compatibility according to the Java SE 17 Language Specification
The following modules define official extension points of the EDC based on the Java Service Provider Interface (SPI), contributing public classes, interfaces and public members which are considered public APIs and are, therefore, covered by the proposed versioning approach:
- spi
- data-protocols/ids-spi
Apart from these SPI-based extension points, individual modules can also contribute additional public-facing APIs, such as communication endpoints (e.g., based on HTTP). To support a fast-paced development of such endpoints without impacting the connector's core release cycle, modules contributing this type of public-facing API can be managed within a separate repository.
The following modules are also distributed as individual artifacts to support a convenient customisation of connectors and are, however, not considered public APIs:
- core/*
- extensions/*
Extensions can in turn specify their own SPI-based extension points. Theses are, however, regarded as internal SPI and not as a public API. Therefore, changing internal SPI doesn't necessarily imply a version increment for the module.
Until its first major release, the Eclipse Dataspace Connector will be developed under the version 0.0.1 without complying to semantic versioning (i.e., API changes don't imply a major release). Snapshot versions may break binary compatibility with previous versions and should not be regarded as stable. There are no guarantees regarding functional and non-functional aspects of the implementation. Tooling for a later migration of current implementations to the envisioned first release will not be provided.
License and notice files must be included in every unit-level distribution artifact. In the case of Java archive (JAR) files, the legal files should be placed in the META-INF directory. However, depending on the distribution format, the exact location of the files might vary.
An appropriate license file is supplied on the root of the source code repository and must be included as is in each
distribution artifact. The supplied top-level notice file represents a snapshot of the dependencies included in all
modules present in the project repository at a given point in time. Before each new release or distribution, the notice
file must be updated regarding the listed third-party dependencies.
While distributing individual modules, a notice file containing only the relevant subset of dependencies must be
created (as described below).
Notice files consist of some prescribed statements addressing trademarks copyright, and licensing. Additionally, the section on third-party content lists all dependencies of the current scope (project or module) and must be maintained before each release. This list is populated by deriving dependencies using the build tool (i.e., gradle), analysing them using an IP tool (i.e., Eclipse Dash Tool), and decorating the resulting report with additional information using a custom script. The shell script located below docs/legal supports parsing the results of the Eclipse Dash Licenses tool and creating a formatted markdown report listing third-party content with extended information.
Execute the gradle task allDependencies for creating an integrated dependency report over all sub-modules of the project (including isolated modules). To process the dependencies of a specific module (e.g., an individual launcher) execute the standard dependencies task:
- First, the dependencies of this module are calculated with gradle and passed to the Dash tool:
gradle dependencies | grep -Poh "(?<=\s)[\w.-]+:[\w.-]+:[^:\s]+" | sort | uniq | java -jar /path/org.eclipse.dash.licenses-0.0.1-SNAPSHOT.jar - -summary NOTICES
- Second, the resulting report is used as input for the shell script:
./generateThirdPartyReport.sh /path/inputFilename
- Finally, the resulting report is assessed for missing license information which must be then added manually.
The Eclipse Dash Licenses tool first looks into IPZilla and second into ClearlyDefined. IPZilla tracks the results (i.e. approved/restricted) of IP due diligence conducted by the Eclipse Foundation. The Dash tool reports for each artifact found within IPZilla also its corresponding contribution questionnaire number (CQ#). In some cases, an approved artifact doesn't reference a license type, which has to be then searched manually. ClearlyDefined is maintained by a third-party and assigns scores to artifact licenses. If a minimum threshold is reached, the item is considered as approved. The Dash tool tags artifacts found within this source accordingly. In some cases, the Dash tool results in an inappropriate license, although a more suitable one is existing. In this case the tool requests a manual review. In rare cases neither an Eclipse approval nor an ClearlyDefined entry is found. Currently, these licenses can be found manually (e.g., on Maven Central).
As far as technically sensible, project modules are packaged and distributed as Maven artifacts via third-party services (i.e., Maven Central).
Execute the gradle task publish on the level of an individual module to publish it as a Maven artifact.
Artifact names must adhere to the following naming convention:
- Group name: org.eclipse.edc
- Artifact id describing the module name (disregarding the directory structure) separating terms by a dash
Examples:
org.eclipse.edc:spi
org.eclipse.edc:util
A comprehensive list can be found here.
Note: the intended audience for this section are individuals who are eligible to author the release process. At the time of this writing these are the committers of the project.
To trigger a new release please follow these simple steps:
- update
gradle.properties
: set theversion
entry to the new version. - trigger the actual release in GitHub:
- on the
Actions
tab pick theCreate EDC Release
workflow - Select the
main
branch - clicking on
Run workflow
should bring up a prompt for the version string. Please enter the version string in SemVer format without any prefixes:0.0.4-something-SNAPSHOT
would be OK, whereasv0.0.4-rc1
would not. - start the workflow
- on the
The GitHub workflow then performs these steps
- creates a tag on the current branch, e.g.
v0.0.4-something-SNAPSHOT
(note thev
prefix). This is done using the GitHub API. - creates a merge commit from source branch to
releases
. The version information is encoded into the commit message. - triggers the Eclipse Foundation Jenkins instance ("JIPP"). This is where the actual publishing to MavenCentral
happens. Note that this process may take quite a bit of time, as every module is signed and uploaded. Important: if
the version string contains the
-SNAPSHOT
suffix, the version is uploaded to OSSRH Snapshots instead of MavenCentral! - Creates a GitHub release including an automatically generated changelog, if the release is not a
-SNAPSHOT
. This is only for informational purposes, no additionsl artifacts are uploaded. The GitHub Release has the following properties:- only created on non-snapshots
- always created off of
main
branch - the release notes are auto-generated based on the last available tag and the
.github/releases.yaml
file - no pre-releases are supported
- no discussions are created
Important: The first commit after a release has to change the defaultVersion
in gradle.properties
to -SNAPSHOT
again. Otherwise, the upload of the automated nightly builds to OSSRH Snapshots will fail.