I’d personally answer yes, but that’s because I believe that typed, schema-generates structures should be pushed all the way from the database to the typescript code on the front end. Changes become something you can deal with with high confidence and things like typos, etc. become impossible. Relational structures enforced in the database (and imho sadly lacking in the intervening layers which is something I’d like to get the time to deal with; though the FoubdationDB record layer certainly has an interesting if unexploited take on it) is a natural extension of this.
But there’s a whole other world where people just don’t want to be tied to constraints. I suspect it’s a personality thing.
Look at Clojure for a good example - it is strongly (but dynamically) typed, but then it also has schemas with automatic validation on top of it (Spec).
I do too. And also probably a lack of experience dealing with outages and remediation of mistakes.
If it's always someone else's problem to clean up your mistakes then you can't appropriately appreciate how valuable it is that a machine can help you make less of them.
incidentally I believe this is the difference between being an engineer and being a product person. I straddle the line. in theory I want a strong type system so I can be confident about the future. in practice I don't want to pay the upfront cost and I'm more interested in shipping now and taking on debt.
In my company's product, we specify the data models as JSON Schema, and then generate the necessary language code -- we generate the GraphQL schema from it as well as Go data types, with some database glue. Our front end code is currently JavaScript, but we hope to migrate to TypeScript, which will make everything statically types all the way through.
(gRPC fills a similar role, though the web story is lacking. gRPC is brilliant for APIs between backend services, but browser support is not there yet. GraphQL is more convenient for the app-facing layers.)
Proto is not elegant and the data model and implementation (especially the very poor performance of maps in the Go implementation) can leave a lot to be desired, but it's a thing that pretty much supports everything. It has good enough performance without headaches. Lowest common denominator can be quite useful and so can constraints.
They don't give too many reasons, but say that operationally it is gonna be easier for them to work with a single database than with two (no-brainer?)
Switching to Postgres made a lot of sense even when looking at Pulp as a standalone project rather than as a part of Satellite.
I work in FinTech and you'll see several backing our apps, and this isn't uncommon.
MongoDB has definitely come a long way from when they were dropping writes and hitting scaling limits, and their new tools ecosystem holds a lot of promise, but I treat them much like any other non-RDBMS store (Redis, Cassandra, Neo4J...) in that they’re very specific tools for specific needs, not something that can be kit-bashed for any general class of problem.