Switched from relying on ID for the blob content to
generated hash-based data-* attribute to make sure we're
generating the unique one for each blob.