Abstract

This is a placeholder for a proposal to add support for Maven 2 to the Roller build process. Currently, it is just a place for Sean to put notes while learning the existing Roller build.

Requirements

There are two basic approaches to adding Maven 2 support, each of which would have a different set of requirements:

1. Maven is the primary (or only) build system
2. The primary build uses Ant and the Maven build is an alternative for those who prefer Maven

For now, this proposal assumes Maven would be the primary or only build system. This means all issues must be addressed to the satisfaction of the team. Allowing Maven as a secondary build system would, perhaps, be easier to implement, but it would create a maintenance issue as the Maven build could diverge from the Ant build. In addition, one of the main benefits of a Maven build, pulling the JARs out of SVN, could not be implemented. So, for now, I am approaching this with a view towards an implementation that would meet all the requirements for building Roller.

Implementation Schedule/Timing

There are a few reasons why the time for a switch to Maven has not yet arrived:

1. A switch to Maven 2 will not add any new features to Roller
2. The Ant build is already using the Maven 2 Ant Tasks for deploying Roller jars to local Maven repository
3. The most active committers on the project (Dave and Allen) are opposed to Maven 2.
4. There are ongoing efforts to make improvements in Maven 2 that may address some of the valid concerns expressed on this list. For example <http://www.jroller.com/mrdon/entry/making_maven_2_not_suck>.

This proposal (when finished) should be reconsidered when two or more of the following are true:

1. A much improved version of Maven is available (e.g. a 2.1 release)
2. The Roller build needs to be restructured anyway (perhaps to break it down into re-usable components, etc.)
3. There are (new) active committers on the project who are comfortable with Maven 2 and are willing to be responsible for the conversion AND ongoing maintenance (including helping the other committers adjust)

(See Mailing list posting)

It may also be possible to restructure the existing build to be a little more Maven-friendly. The Ant build could use the Ant Tasks for Maven or Ivy to pull in JARs, for example. This could allow some of the benefits of Maven, without all the pain. This could be a transitional step towards a full Maven build.

Issues

General Issues

• Maven is controversial, some love it, others hate it
  • We'll need to clearly lay out real benefits and drawbacks in this proposal.
  • Hopefully, the Maven team will make improvements to Maven to address valid complaints.
• Maven allows simple builds to be implemented "declaratively" but mechanisms for implementing custom build logic can be frustrating.
  • We'll need to make sure all current and future build logic can be handled by Maven.

Specific Issues

• Roller Build Advanced Features (A list of things in the Roller build that would require a Maven plugins or scripting should go here)
  • Multiple sub-projects and build artifacts mandate a Maven "multiproject"
  • Customized Builds: Current Ant build can be customized for a specific installation using Ant XML fragments. (How would this mechanism "port" to Maven?)
• General Maven Issues
  • e.g. XML verbosity go here

Benefits

It will be important for these benefits to be clearly described and contrasted with mechanisms in the current (or an improved Ant build.)

• JARs can be removed from SVN and stored in a single location on a developer's system
  • Speeds up SVN checkouts
  • Reduces storage requirement on developer's system (especially if simultaneously using multiple branches)
  • Note: This could also be done wit Ant tasks for Maven or Ivy.
• Maven plugins can be used to generate IDE project files
  • To generate Eclipse .project and .classpath, use: mvn eclipse:eclipse
  • To generate IDEA project, module, and workspace files, use: mvn idea:idea
  • What about NetBeans?
• Standard Maven layout and project site will make things easier for experienced Maven users
  • This is a little bit subjective, but clearly some will see it as a benefit

Drawbacks

(How are these different than "General Maven Issues" above?)

• Effort of Implementation results in no new features for Roller
• Maven can make custom/advanced build features difficult to implement (see list of issues above)
• Maven may not be already installed on Developer's systems

Comments

Please enter comments below or send comments directly to Sean Gilligan – we're trying to hold off on discussing this on the lists for a while.