To implement the object concepts, Netsilon works out an object-relational mapping whose semantic is particular. Each object is identified by an object id whose column is found in each table. An object identifier carries also an identification of the class of the corresponding object. A particular table, instances, is at the same time used to check the class of an object and to find the whole of the definite objects. An other table, reference, when it exists, is used internally by Netsilon to store information needed in the incremental modification process of a database schema.

Each deployment site has a specific object-relational mapping. Netsilon converts the name of the model elements to a name valid in an SQL context. A name of a table can be prefixed if a prefix is defined in the database properties. During incremental schema modification, Netsilon only considers the tables whose name starts with the prefix. It keeps the name it first gave to a model element weather the name of this element is changed or not in the future.

Netsilon authorizes the modification of the naming of the tables and the attributes in the generated schema and has a command to edit the mapping. The table name should be prefixed with the prefix specified in the database properties or else Netsilon will recreate the table during the next deployment. The modifications made to the mapping must also be made manually to the schema before the next deployment or existing data will be destroyed.

When a class is deleted from the business model, it is of primary importance to let Netsilon update the data base under penalty of obtaining null objects during execution. If the database is not accessible on line, a local copy can be made, updated and reinstalled on the site.