There is a difference between the rewriting method that appears in the ER 2013 paper and the method presented in the MSc of P. Manousis: this difference concerns the case of an Attribute Addition to a Relation, in the case of a veto. We believe that this improvement in the case of ER 2013 correctly reflects what would actually happen in a real case.

In both cases, if the relation itself vetoes the addition, the veto is respected. The difference lies in the case of a veto by a subsequent module.
  1. In the P. Manousis' MSc, a veto to Attribute Addition on the raltion module, by a subsequent module would block the addition to the relation too.
  2. In the ER 2013 paper, however, we actually add the attribute, despite the veto of the subsequent module. Of course, all the modules that veto such a change will keep their original schema, ignoring the Attribute Addition. Moreover, in order to be semantically correct, if there is a Group By in the module that is to be rewritten, we add the new attribute (say 'new_attribute') to the output of the module via the aggregate function count (new_attribute).

This difference incurs higher rewrite costs for the ER 2013 case, as all the modules that accept to include a new attribute will be rewritten.