Because the second can more easily be parsed 100% correctly and consistently, while the first one will not allow you to fully understand a document-tree without

Because the second can more easily be parsed 100% correctly and consistently, while the first one will not allow you to fully understand a document-tree without lots of wile and complex forward and backward document scanning, and you may still be wrong. This especially true for bigger and deeper document-trees.