I'm looking at a design specification of a work request object that would if implemented purely as a standard salesforce object would cause very sparse columns requiring tens of recordtypes and hundreds of heterogeneous hidden fields.
I'm considering Abstracting out many of the sparse fields into child attribute-value records to be creating by deep cloning 'template' entries and viewed and updated by custom VF pages.
I should be able to correctly assign types the attribute values by giving the the AV object multiple record-types for which the correctly typed value field is the only one visible.
Has anyone tried this EAV approach on SFDC? Or have any suggestions on how to manage this sort of complexity otherwise?