> what makes kdb so special and why isn't there an open/libre alternative?
Maybe I'm reading too much into this, but it seems like you expect the answer "nothing much" to the first part of your question. I have done nothing more than read lots of articles on KDB's lineage and play around a bit with J, so take my answer with a considerable lump of salt, but my impression is that the answer to the second part of your question is "because there's something 'so special' about KDB"; my understanding is that it provides blazing-fast access to memory-compact databases, from a tiny codebase, building on the APL/J legacy. Why can't this be done in an open sourced way? Well, surely there's no inherent reason, but the fact that it hasn't been is probably evidence that it's not just a problem of trivially cloning existing work (or else someone would have done it).
The real answer is Arthur Whitney is a god-like being with coding powers beyond the mortal realm. Seriously though they're a lot of stories of his excellent work if you search for them. He's supposedly working on kOS so you can run kdb+ on bare metal. Tiny, fast code is what he does. Sure it might look incomprehensible, but it is a few pages of code he can store in his brain at once. Aaron Hsu (someone has some links to his HN posts on here) explains this with his compiler which converts APL to GPU code. His entire compiler (which he has worked on for a long time) is a handful of pages. He says something like "there's no need for an abstraction if I can see everything at once". Most programming languages have implementations that are very long...which do you think has less bugs? Steve McConnals Code Complete has a statistic somewhere on bugs per 100 loc. That doesn't leave a lot of room in Arthur's code base (although you could say his code being so terse makes it an apples to oranges comparison I guess).
Maybe I'm reading too much into this, but it seems like you expect the answer "nothing much" to the first part of your question. I have done nothing more than read lots of articles on KDB's lineage and play around a bit with J, so take my answer with a considerable lump of salt, but my impression is that the answer to the second part of your question is "because there's something 'so special' about KDB"; my understanding is that it provides blazing-fast access to memory-compact databases, from a tiny codebase, building on the APL/J legacy. Why can't this be done in an open sourced way? Well, surely there's no inherent reason, but the fact that it hasn't been is probably evidence that it's not just a problem of trivially cloning existing work (or else someone would have done it).