Absolutely. I’ve been going through this tech research process now as I need near-real-time or at least reactivity and I just scratch my head and think “why is this so hard? don’t everyone need this stuff these days?”
> This is non-trivial in the relational model.
Yes, but I don’t believe that it has almost anything to do with the model, and almost everything to do with choices in RDBMSs like Postgres, which are slow moving tankers.
All the abstractions and procedures for materializing views (ie store and run the queries in a push instead of pull fashion) are both theoretically and practically well understood, and half of the support is already there.
WAL + CDC is virtually already the standard, it’s just clunky to use. The missing pieces seem to be more about multiplexing subscriptions and it’s implications on downstream protocols (probably needs wire protocol changes?).
In either case, I very much think the relational model should not be considered mutually exclusive from reactivity, and the implications of assuming that may be throwing out millions of holy babies with a little bit of bath water.