imho, XDoclet1.x tags are javadoc tags. So it has to be specified in a
java source file. The parser will not be able to parse a properties file and extract metadata.
XDoclet2, ideally yes it was'nt supposed to care where the metadata came from. So even though the design is a lot cleaner when compared to 1.x, its not there yet.
May be you can have your properties file in the form of XML with the required attributes and then get
ant to apply a stylesheet on it and generate HTML and .properties file as well?