It shouldn't be imho. I think it's incorrect and is a result of bad analysis leading to bad table design. I went on a course years ago where we discussed this and said that you *might* consider making that a full join and creating a context with only that one join in it if you have denormalised dimensions in one of the facts but not the other. However, you'd only do that if you were certain of the granularity of the join. We agreed, though, that the model was principally flawed and it would be a lot less painful to reassess the model and rework it.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.