The fourth layer is created by a more granular splitting of the 3-tier architecture. In my case the first layer is splitted into two layers. The first layer loads the raw-data without much transformations (adding some fields for incremental approaches, some if-constructs to handle different formattings for the same source and few other things but not really changing the data itself). The second layer used the qvd's from the first layer to do all the transformations. My third layer then creates the real datamodels by combining various qvd's from the first/second layer and the fourth layer is just the binary loaded dashboard with the gui.
If I would consider the logic of using layers more strictly I would need even two or three layers more but I'm a bit pragmatically on this point because most of my data needs only 3 or 4 layers.
In my scenario I load the section access table within the third layer (datamodel) and the binary load transferred it like the other data into the dashboard (fourth layer). In both layers I enable the section access settings within the document properties in tab open.
I must admit that I use this since many years (without thinking much about it) and I'm not absolutely sure if it wouldn't be working to load the section access just after the binary statement and to keep the section access only within the last layer.
Section access is a very special feature and you will need in each case various testings if it worked for each user like expected. Therefore I suggest that you just makes some more testings in which layer it would suit best. But before applying any section section make sure that you have BACKUP's from all of your applications - otherwise could be a mistake very painful.
- Marcus
