Relax already ;-)

In response to Craig, I should clarify my stance on Relaxer. I know several of the articles on Relaxer (including the site) focus on its use as a data binding tool. Haven't used that. I don't use it for data binding (or reading schemata at all), rather I use it to create XML Schema, DTD, and XSLT stylesheets.

In particular, I've had good results creating XML Schema using Relaxer, tweaking a few things, and then running it through JAXB to create my data binding libraries. (Even on some real world projects!)

XMLBeans looks promising. I've yet to use it on a project, but I particularly like that they have access to a Java model, an XML Infoset model, and a PSVI (and that they are synchronized). If I used XML Beans, I would probably still use Relaxer to begin writing my XML Schema (unless I already had one). That's because XML Beans is a "start-from schema" binding tool, like JAXB. They are considering adding JSR 101/109-compatible "start-from Java" binding. Relaxer offers a start-from XML approach that can get you a W3C XML Schema if you don't already have one.

I have hit some glitches in Relaxer, but I've had enough success with the beta that I was excited by the 1.0 release.