Ok, so I understand the new join datatype that allows a 'parent/child'-like relationship between different documents in an index, however it seems to me that these documents still need to all be the same type (due to the (also new) single types in each index rule). Does this mean that we are now limited to only ever have 'parent/child'-like behaviour between documents with identical types in an index.
To elaborate on my usecase, I am storing a number of sensor readings which are also grouped by a single event. So before my parent would be 'event' and my child would be 'reading'. This is one to many, generally one event has between 250 and 500 readings. In 'old ES' I would have an event type and a reading type in the same index, linked by a parent/child relationship. This would allow me to store other information about the events that relates to all of the child readings.
It might be that an old style parent/child between different types in the same index wouldn't have been the right way to do it anyway, but I'd certainly welcome a pointer as to the best way of doing this (even if that is to just denormalize the data)
Thanks in advance