Check document bodies contain mandatory fields, instead of just checking for emptiness.

1. in one of the cases it was redundant to check for emptiness before checking the fields anyway
2. checking for emptiness does not make sense, as there are a lot of 'errors' it does not catch in non-obvious ways: a) xmlns attributes count as non-emptiness b) unknown tags/attributes count as non-emptiness

---

Migrated from D4469 (view on Phabricator)