I have realized something which has caused a change in thinging. XBRLS really should be two pieces, not one piece. Right now, XBRLS combines an architecture and a modeling layer. There are benefits to having these two pieces separate.
The benefit of an architecture layer is that you can constrain the generally too numerious options offered by XBRL. You can also remove unnecessary and undesirable flexibility which exists within XBRL, making your system easier to build and use.
The modeling layer helps build your taxonomy consistently, helps users of the taxonomy understand the taxonomy easier, allows for software to be created more easily and enables the creation of features which make it easier for a business user to create, use, and report against a taxonomy.
While all systems need some sort of architecture and some sort of modeling layer; separating the two allows for reuse of the architecture. I am not sure if a modeling layer would be usable if it were moved to a different architecture, have not thought about that much and have not really experimented with this idea much.
More information on what a modeling layer is, its benefits, and how to create one coming soon! Stay tuned.