Hi Josiah Haswell,

Just so that I get my thoughts in order, you cannot use the the current strategy because attribute names are prepended with the table name in Datomics? Presumably to give them proper namespace isolation. By using the joined strategy, you would cut down the number of attribute metadata created. Is that correct?

If yes it seems to me that a EAV database might prefer the SINGLE_TABLE strategy where all columns are stored in the same table. The EAV model is after all a way to map sparse matrices which is what SINGLE_TABLE models are about. am I missing something? Is it because in EAV/Datomics you purposely do not want to have discriminator attributes? In which case, how do you know that a given entity id is both persistable and foo (to know it is in the end of type foo)?
I don't know Datomics much so I might be completely off base.