thanks ilja.
you are right, that class-model and mapping must not be close. but to make things better comprehensible, i try to get a mapping which is not too far away from the domain-model. especially when not only restructuring domain-model but adding new attributes to new evolved domain-classes and squashing them to an old o/r mapping things could be very difficult after a while.
so i think the mapping should evolve with the domain model too, especially in the beginning and some effort should maybe put to a kind of automatic refactoring domain-model ETL tool for the already existing data sake. but haven't tried out yet, still seeking for more (and maybe better) ideas
thanks for the link too. i had a rough look but at the first glance it did not mention the handling of an evolving domain-model with an o/r mapper. it merely went more into detail when it came to structure of relational databases.