Thanks for your reply, Michael!
Do you intend to create a separate discussion forum for Antora? Or is this the place?
Great question. We hadn't thought about that yet. For now, I'd say we can discuss here, since this project is very closely connected to the Asciidoctor ecosystem. Once the project matures a bit more, we'll likely look into having a dedicated forum and chat channel.
As a use of a dependency management system is familiar to most readers, maybe you could draw the parallel (and point out the differences) in a future blog post?
One difference I am reading, is that in software development, the component is responsible for building (publishing) its re-usable artifacts.
What you're describing is exactly how Antora is designed to work. The current content aggregator happens to be looking for files in repository branches right now. But there's nothing preventing it from taking files from a published jar file instead (or in addition).
Where content is stored depends on the workflow of the organization. Some organizations use repository branches instead of published documentation artifacts (as they consider the latter to be too much process). Others publish artifacts, as you have suggested. Neither is right or wrong. It's just what works best for each group.
The cornerstone of Antora's pipeline, which will be covered in an upcoming article, is the virtual content catalog. After the content aggregation step, no subsequent component should touch the physical file system (or worry about fetching content). All the content should be in the virtual content catalog. This abstraction allows a myriad of choices as to where content is sourced. It could be a repository branch, a local file system, or the contents of a published archive. As we emphasize in the first article, content is sovereign.
Antora is proposing that the generator goes into the components repository and fetches what is needed. I feel that this blurs the demarcation of responsibilities, although is saves the need for an artifactory of some kind.
Antora is simply proposing that the generator collect content from disparate sources. Which sources, and how those sources are managed, is a choice the organization must make for themselves. Antora should support that choice, no matter what it is. We think using repository branches is the easiest and most elegant way to get started, but Antora can accommodate other arrangements too.
I love the idea of a standard project structure for documentation. I hope this also results in guidelines to address issues like https://github.com/asciidoctor/asciidoctor/issues/894
Indeed.
Thanks again for the feedback!
-Dan