For the record: I disagree that this is an enhancement. ElementTree supports PIs as first-class tree elements. They can be added, inspected, removed, and written out when serializing into XML. Only when reading in XML, they are silently dropped. I think this is a bug, no matter how rarely people might stumble upon it in practice.

If there was no support for PIs at all (i.e., they couldn't be created as tree objects) I'd agree with you that this is an enhancement.


Unless there are objections, I'll try to work on a patch that either documents that PIs are lost, or optionally adds them to the tree when parsing (depending on how difficult that turns out to be).