best XML aproach

Trying to find the best aproach to an XML parser, I end in the follow 3 situations: 1 - We want to parse an XML that we don't know nothing about it 2 - we want to parse an XMl that we know how it can be, but we don't know how exactly it is 3 - we know exactly how it is. let's use the follow xml to try all the 3 aproach above <family> <father name="Marcos" age="23" /> <mother name="Patricia" age="26" /> <childrens> <paul>I'm Paul</paul> <mary>I'm Mary</mary> </childrens> </family> 1) using the first situation, as we don't know how the xml is, we need a class that implements the follow methods $class->parse($xml); // build from a xml string or file $class->getAtributes(); // an map array with the atributes $class->elements(); // the same as above $class->getData(); // return the CDATA this way, we can walk throw the tree. XML_Tree uses this aproach (I think!) 2) The second situation, that I thing is missing in XML_Unserializer and others classes. the class would be like this: The first method we set an schema for the class, so we could still have an empty xml, but with all properties like this: $schema = " <family> <father name="string" age="number"> <mother name="string" age="number"> <childrens></childrens> </family> "; $class->setSchema($schema); $class->parse($xml); the data would be $family: object $family->father: object $family->mother: object $family->mother->name and ->age // the same for father $family->childrens: array you see that even if <childrens> is empty, has 1 or more, or even inexist, we still have the childrens vector, thanks to the schema XML !!! 3) this aproach is what XML_Userializer. Since childrens data type is always changes dependent on xml. I Hope this help... I preffer implement new features in an existent package instead of adding new one to the repository.