Michael,
I didn't catch the opportunities for reuse from your first post, but I think I got it. It does sound like a no-brainer that giving people a DTD to publish to, and having a single import routine for that DTD, would go a long way towards easing data-model learning curve/migration issues for a OpenACS newcomers. That's something I would definitely find helpful in approaching a new system.
FWIW, here's a project I discovered a while back but haven't looked into that appears to reflect a lot of thinking on this topic:
XML-DBMS
It's a perl port of a Java project to perform this mapping between persistence formats. I'm not suggesting it as a platform, just that it seems well documented and potentially of interest.
