← Back to context

Comment by zooq_ai

9 months ago

tl;dr -- Rust is not the language for startups who want fast iterations and are still finding it's pmf. It is perfect for rewrites of popular system software which already has millions of users and running in production because once a software becomes mature and at million user scale, safety and security becomes paramount

> for rewrites of popular system software which already has millions of users and running in production

Hopefully, it's perfect for more than that.

Rewriting a system that's already acheived that scale is a contentious decision, to say the least. Many of our colleagues have been lost to and scarred by well-meaning attempts to rewrite those systems in the past, including many who were sure it was the right choice for "safety and security" themselves. For every conceptual safeguard new tooling might give you, you invite countless and sometimes catastrophic regressions in actual application logic in writing new implementations. It can work out, but no tool can expect to live only on that kind of work.

I'm sure Rust has much broader applicability than that. Or I hope so at least.

At million user scale, feature parity and compatibility is truly paramount and, `rg` aside, Rust utils don't have a great story to tell there. Such software has evolved in the course of decades and its (usually despair-inducing) code is tailored for dealing with a tremendous number of edge cases, platforms and architectures. Rewrites in Rust tend to write "the cool parts" and indeed write them better and faster, but that is just not good enough for replacing the standard tools.