Real Story Group. Make Better Technology Decisions.

Kas Thomas

JBoss DNA: using JCR to make metadata behave

22-Mar-2008

Tags: Enterprise Search, , Information Architecture, Red Hat: JBoss Enterprise Portal Platform

The JBoss folks have launched an ambitious new open-source project that is so breathtaking in scope, it defies easy categorization, even though (ironically) it is largely about categorization.

JBoss DNA is (according to the project website) "a repository and set of tools that make it easy to capture, version, analyze, and understand the fundamental building blocks of information." Notably, the key enabling technology for the project is the Java Content Repository specification (JSR-283).

The project description goes on to say: "As models, service and process definitions, schemas, source code, and other artifacts are added to the repository, JBoss DNA 'sequences' the makeup of these components and extracts their structure and interdependencies. Users can then search, analyze, visualize, report, and modify the repository's content using the terminology and structures they are familiar with. Such domain-specific solutions can be created with little or no programming. Sharing this information is possible through Eclipse plugins, web applications, and REST servers."

If you're still not getting it, there is a useful slide show on the JBoss Labs site. The system is transactional, event-driven, and rules-based, and (of course) it leverages a long list of well-known open-source building blocks and industry standards. Basically, what it does (if I understand it right) is allow you to discover and manage dependencies and semantic relationships between bits of info that most of us would otherwise call metadata.

A core primitive in the JBoss DNA system is the Sequencer, which is essentially a custom event handler that fires when you insert a content item into the repository. It executes rules (which you write in a domain-specific rules language) against the item in question, to extract atomic bits of information about it. In other words, a sequencer does autoextraction of metadata. (Why don't they just say that? Why the cutesy bioengineering lingo?) According to the project's leaders, sequencers are planned for .zip archives, Java bytecode, WSDL, UML, and database DDLs, among other targets.

Autoextraction of metadata is a noble goal, of course. In fact it is becoming a key capability in many corners of the content-management world (DAM in particular). But there are problems with the JBoss DNA vision, not least of which is the fact that metadata extraction is notoriously tricky business (and inferring taxonomic relationships gets even trickier). Visualization of this kind of information is also challenging (ask any Edward Tufte fan), a subject on which JBoss DNA is silent.

But the greater issue with the JBoss DNA project is that the problem space, as envisioned by the project's creators, is hopelessly broad (King Kong could not get his arms around it) and the DNA reference architecture is bewilderingly baroque, encompassing federation of repositories, a Publishing Server that implements the Atom Publishing Protocol, WebDAV support, pluggable analytics, connectors of all kinds, and scads more. The number of moving parts is large and the footprint will doubtless be massive. The factoring is distinctly J2EE circa 2005, in the most obnoxious sense. If there is one thing IT departments don't need at this point, it's yet another kitchen-sink Java EE architecture to deal with.

Earlier, I referred to JBoss DNA as ambitious. Perhaps audacious is a better word. To be sure, many sublime achievements in this world began as audacious dreams. But it is true, also, that audacious endeavors sometimes (maybe most of the time) end up as giant, smoke-filled craters; and JBoss DNA, for all its noble goals, already seems in danger of following that trajectory.