its kinda like this but more complicated, as I have various resources
like bookmarks, files, contacts etc, they are identified by various
unique compound keys. if I'm understanding the approach Mongo is
using correctly, we should only pick the most common fields in various
docs and make them unique, as a record does not have the field which
meet the unique compound key constrain, say the key is
"serial_-1_type_-1_owner_-1",
and a bookmark record does not have "serial", Mongo recognise it as
"serial=null",
so I will get duplicated error if I insert another bookmark record,
correct?