Here is a little bit of extra information.
When I made the xml_archive I considered a number of options. I settled on the simplest implemenation which I thought would be considered acceptable. I also tried to use map xml features such as object id tags to analogous concepts that the serialization library required. This has worked out pretty well in that not too many people have complained that its not enough.
I did consider have the serialization library generate a compatible schema when the archive is saved. This might be useful for browsing the archive. I didn't do this, but it's still an interesting idea. I could be implemented by making an new xml_oarchive, by deriving from the existing one or by making a whole new arechive - xml_oschema.
Good luck.
Robert Ramey
"Joseph Fradley"