okay, so i feel trapped.
for me, spider web kind of attribute relationships looked suitable for a number of reasons. namely:
1. member properties are created from attribute relationships. so if i want my product leaf item to have many levels of product groupings as member property, what can i do? either i add redundant relationships. or i make copies of the attributes, and use one copy in hierarchies and one copy as member propery:
K -> A1 -> B1
K -> B2
B1 and B2 being the same physical data, B2 has no attribute hieararchy
2. adding redundant relationships provide extra information. sometimes there is a flexible relation from A to B, and from B to C, but there is a rigid relation from A to C. i thought to be kind, and tell this information to AS.
however, it seems that AS is not grateful at all. more precisely, it says:
"If a measure group exists where the granularity of the data is at a level other than the key attribute (attribute A) level. Attempting to aggregate data from the key attribute A to level C will not return the expected values. " ( http://support.microsoft.com/kb/2127437 )
oops. what now? should i go the roundabout way, and add great many "fake" attributes to provide 2nd 3rd etc level member properties? or this is just "fear mongering", and my cubes will work just fine? doubt so. not that i understand what the actual danger is. "attempting to aggregate" when? who? it affects predefined aggregations? or query-time aggregation too?