Description
On a large commercial DFDL schema project, the XML documents being produced by the parse have truly excessive numbers of entirely unnecessary namespace bindings in them.
The file size of XML infoset files is dominated by the namespace bindings.
Technically the XML is "correct" and tests don't fail, but the volume of these bindings has first order impact on:
- developer productivity - can't see what's going on in the XML data because it is so cluttered with bindings. What's worse, pretty printing gets very indented and ends up inserting whitespace around simple values (which breaks comparison) just due to these bindings.
- performance - the XML output is vastly bigger than it needs to be. XML is already big and slow, but this is just piling on!
If this fix is reasonably localized, I might request we make a patch for 3.4.0 (ex: 3.4.1) as this issue is very important.