There’s always battles waging in the technology community – Linux vs Windows, OpenStack vs AWS, DevOps vs NoOps vs SomeOtherKindOfOps. Another oft visited argument is the relation versus non-relational databases. Essentially the protagonists for the NoSQL movement suggest that relational databases are too slow, and don’t map well to complex data patterns. Antagonists counter that only a select bunch of massive users have the database size that makes performance a problem, especially since modern relational databases are faster than ever before and that mapping issues can be somewhat resolved using different approaches.
Of course more moderate folks, without a dogmatic view one way or the other would say that for some applications a relational database approach is preferable, while for others NoSQL rules the roost. For several releases now relation database PostgreSQL has begun expanding beyond a rigid relational/non-relational definition with a key/value store, the JSON datatype, and GeoSpatial support. The idea being that this approach makes a single database more flexible to different data approaches than previously. Essentially this is a case of seeing a database as a platform onto which extending technologies can be brought.
The key benefit here is one of aggregation This move isn’t earth shattering but it brings a more complete feature set onto the one platform. It’s a case of backfilling functionality that reduces the likelihood that heroku developers will look elsewhere for solutions. There’s lots of alternatives out there, but for existing Heroku users, the platform just got a little stickier.