Comment 5 for bug 580526

Ryan, so, the reason to go with the "standard" XML support in intltool would be because of all the bits and pieces that it already provides (like xml:space="preserve" support). If we do decide to not care about that in gsettings (we very well might), we should make it clear.

Now, comments on the specifics.
 - use XML::Parser should probably be done in a similar way it's done in readXML (it allows to use intltool without XML::Parser if you are not using XML features); or, we could just drop the eval-hack inside readXML instead and require XML::Parser in all cases
 - yeah, those "my" declarations do look nice ;)
 - it'd be so much easier to work with this if you added a testcase that covers all the important bits; I know intltool test system is far from perfect, but it's still very useful for tracking regressions and understanding how stuff works — especially since I have a hard time finding what an actual gsettings file looks like ;)