It does not have multi-master replication natively, though there are companion projects that enable it.
It does not have multi-master replication natively, though there are companion projects that enable it.
Adding extra implementation to application, to do the database's work against the usual division of labor, no thanks. And bringing some shady 3rd party support package on top of the database is a second really big no.
For businesses only what you get out of the box counts. Otherwise you can always state that "the platform has C++ compiler, stfu and code". That doesn't really work. The companies are into buying complete products for a reason: they don't want to implement unnecessarily things they really should just get out of the box.
TBH I love postgres, but before they offer proper multi-master transactions out of the box most companies just still go for Oracle. That's similar as to why people buy Photoshop. GIMP fell some 10 years after in usability and features when they did not prioritize implementing the dynamic layer effect system. Instead of that they chose to masturbate with GEGL or something like that for 10 years, without providing the actual end user requested features. The situation still stands there, before they implement what Photoshop 4 or so did they got no chance of being valid alternative...
Postgres is my preference over MySQL/MariaDB, but I wouldn't have pushed for it back when I worked in traditional corporate IT: Stability, and having a third-party to blame, really wins there!
http://blog.2ndquadrant.com/bdr-is-coming-to-postgresql-9-6/
Also, we're only on 9.5, so I haven't gotten to play with BDR yet, though I do have a meeting next week to start talking about how we might use logical decoding.
The other thing is that "PG 9.6 brought ..." might suggest this is built into PostgreSQL 9.6. That's not quite the case - we're working on moving it into core, but that takes time. For now it's an extension (plug-in). PG 10 will however include pglogical, which is another step in that direction.