We could do a lot better than SQL. It's just that nearly every replacement falls into one of two traps: A) Not being able to express relational semantics. In

We could do a lot better than SQL. It's just that nearly every replacement falls into one of two traps: A) Not being able to express relational semantics. In general, every programming language replacing its predecessor allowed to express about the same semantics, even if it wasn't natural to the new language. One can write imperative Java/C++ even if the language doesn't like it, and successful functional languages allow escape hatches for mutable objects. The various NoSQL languages typically fail this hard. B) Not offering enough of an improvement. Minor improvement isn't worth the 'yet another query language' burden. EdgeQL doesn't fall into A, but it may fall into B. NULL is an annoyance, but not big enough to justify another language. Throwing an exception instead is a very double-edged sword. The author needs to show far more improvement to justify a new language (I'd have liked to see more examples of composability for instance).