Just to clarify what I'm saying, again: 1) I do care about the use of the translation of a RDB in RDF: namely, query answering should give the same answer in the original RDB and in the translated RDF graph. 2) To capture SQL NULL values, it is impossible to just rely on a translation of the data without considering also a translation of the queries. This means that the translated data alone does not / can not capture the real meaning of SQL NULL values. So, the only thing that makes sense is to consider the whole process of translating the data and the queries as defining the "mapping". 3) I understand that people may not like this, namely the fact that the data alone may be meaningless without understanding how to access it properly. The only alternative is to avoid the translation of SQL NULL values. 4) In the Wiki you find why the translation which forgets NULL values would be wrong without associating to it also a way to translate queries (which can be found if in this case schema information can be accessed). --e.
Received on Thursday, 19 May 2011 20:31:09 UTC