I’m a little underwhelmed by the article. It doesn’t tell me anything I couldn’t have found out on the Rust home page.
I know why they chose it, but I’m far more interested in what they think about it now after two years of actually using it. It’s not a, “Rewriting TiKV in C++” post, so presumably it’s going well? Has it met the team’s expectations? What were the pain points? How is the tooling versus what they were used to? How fast was new development vs C++?
They didn’t want to use a certain type of database because they decided it was immature. That’s a good idea: data is often the lifeblood of a business, so you don’t play games with it.
Then they decided to go with Rust in 2015. I am looking forward to using Rust for professional projects in a few years, but the ecosystem still has a lot of bleeding to do in 2017. Ecosystem matters.
So I think their reasoning has a whiff of rationalization to it.
But someone has to blaze that trail for the rest of us, so hey… have at it TiDB dudes.
What aspects of the ecosystem need to be more mature to implement a storage engine? It seems so domain-specific that one would be writing a ton of their own code anyway.