This looks like a major braking change as probably nobody supplies our DTD.

@Godin Do you really think this optimization is worth it?

@marchof

probably nobody supplies our DTD

what about inlining of DTD into XML:

<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<!DOCTYPE report [
  <!ELEMENT report ... >
  ...
]>
<report>
...

or maybe even XSD ?

looks like a major braking change

The second part of description not implemented here about change for attributes marked as required is indeed a breaking change.

The first implemented part about change for attributes already marked as optional since a very long time did not looked as breaking to me, exactly because consumers already must not assume their presence.

Unfortunately you're probably right - no one reads DTD as like any documentation. SonarQube for example assumes presence of these of attributes 😞

worth it?

14% just for our small project and just for first implemented part, and 27% in total with second part...

I like the idea of inlining the DTD! Also we can improve our DTD to define default values. Then existing parsers shouldn't be affected.