感谢回复
1.既然如此，那和Kimball的dimensional modeling建立的data mart有什么区别呢？
2.在物理层上不做约束，只是逻辑约束？
3.CIF一般好像都采用增加一个时间列和原来的主键共同作为新的主键，是这样吗？
4.你说的用parameter maps，具体如何使用。我看到一篇文章中有如下的方法：
If the mart needs to supports different languages for different users, some database system (such as Oracle) provide tricks to easily create transparent multilingual environments. In Oracle, a single database instance can support multiple schemas. It also allow you to create synonyms in one schema to reference a table in another schema.
As an example, let's say you need to create a data mart that must support both English and French users. Within Oracle you would create three schemas:
l One to hold non-language-sensitive data, such as fact tables
l One to hold English language dimension tables
l One to hold French language dimension tables
The dimension tables in both the English and French schemas would use the same table and column names and the same primary key values so they appear to be same to the query tools. The language-sensitive schemas would also contain synonyms pointing to the fact and other tables in the common schema.
The data delivery process would simply create two versions of the dimension tables, one in each language. User accounts would be established with either the English schema or the French schema as the default schema, based on the user's language preference. When the user logged into the database, the user would be presented with dimensional data in the language of choice.
建立了多个schema之后，如何在具体的工具，比如Cognos中应用呢，做到用户登录后就可以看到自己设定的语言的内容了，是数据库中的内容即报表上显示的内容，不是工具界面上。 |
