I've opened a pull request containing these changes. Robert, if you agree with this, go ahead and merge it. On Mon, Mar 13, 2017 at 11:08 AM, Chase Tingley <
chase@spartansoftwareinc.com > wrote: Currently we model originalData as an array of objects that represent an id and a corresponding data item: "originalData": [ { "id": "d1", "data": "[C1/]" }, { "id": "d2", "data": "[C2]" }, { "id": "d3", "data": "[/C2]" } ], This is following the XML representation, but originalData is really just a lookup table, a structure XML doesn't handle well. However, JSON does, so for JLIFF we can use a more natural JSON representation by treating the id value as an object key: "originalData": { "d1": "[C1/]", "d2": "[C2]", "d3": "[/C2]" }, This is more concise and also easier for a JLIFF consumer to work with. (Additional note, the value here would need to become an object to support the optional @dir or @xml:space metadata supported in XLIFF.)