Thomas Mortagne: This would also allow starting a migration from an existing extraction without having to parse the confluence package again if we also save the state of the ConfluenceXMLPackage class. This can be easily done by dumping the non static and non injected private fields of the class, which is easy to do with the writeObject / readObject feature of Java, which has the downside of not being a very readable serialization format (binary), but for which tooling probably exists as it is a widespread format. Apache Configuration would need support for this and this is not easy to add, but we could use JSON, but this requires adding dependencies for just this feature.

Starting migrations without having to extract the package again would be a huge improvement allowing users to extract select spaces from a full package, and then other spaces in future migrations way more quickly. Full packages are better to extract individual spaces from because they allow links to other spaces not to break.

WDYT?