This is actually a blog post I have been meaning to write for quite a while not, but I just never got around to it. Relational databases are ubiquitous in programming. Whether it’s mainframes, application programming or even phones now a days everyone stores persistent data in a database. Recently there has been some talk about non-schema databases and there’s a myriad different implementations of them, but why the sudden interest?
- Database has at least two things that makes it less than ideal for agile development. First one being types, especially if the types propagate out into the application code (*cough* Microsoft). The second being based on a schema, which means that one has to make all the choices up front.
- Databases are really built for a single machine, and even though all the big database vendors has replication support, it still seems like a bad solution if one really wants scalability. There are a lot better ways of doing scalability now a days (a distributed hash tables comes to mind).
- Databases are meant to be standardized, but in reality they are not. They all have different syntax to do common things.
- It has been said that when you build a DSL, then sooner or later it will turn into a (bad) full-fledged programming language. What if instead the data was simply part of the program and you could manipulate it as you are used to? Functional languages makes this much easier and possible. Especially a programming language that is built with concurrency in mind.