Null References: The Billion Dollar Mistake - Tony Hoare

There's a constant tension between bit-packing and rich semantics older than me! As you likely know C.A.R. Hoare calls `null` his "billion-dollar mistake": [this video] . I linked the Cliff Click talk because it's the best example that I know about of someone using some semi-formal rigor around valid states and the transitions between them in the context of raw atomic operations to get a quite clearly correct result with a legitimate performance imperative: he had a machine with 768 cores, customers who bought that machine to have 768 threads jump on a hash table all at once and wanted ~~700x more throughput having paid for the machine. But even Cliff Click made sure that both his conceptual states and their physical representations were going A -> B -> C w.r.t. a given conceptual "register". I'm not trying to knock the serious concurrency pros who need atomic and/or relaxed operations in some super hot path: there are unambiguous use cases for this stuff. I've legitimately needed these things once or twice myself. Lock-free is a bad default. Even the elite concurrency pros work very hard to get something that's both robust and still a win after the constant factors. I'm saying "proceed with high caution, and on the basis of measurement", not "never ever do this".

> Sure. With infinite energy, anything’s possible - we can prevent all bugs. The problem is, we don’t have infinite energy. You don't need infinite energy but it is a significant undertaking. The UB seem to obey a power law in regards to their lifetime i.e. every X years number of UB caused problems halves. > I can only recall one serious concurrency bug I’ve ever had (a data race) which took some debugging time to track down, but it wasn’t in Go. YMMV. That's what is nasty about UB in data races. It's not trivial to find and it's a pain in the neck to reproduce. So even taking you at your word, you not having issues with data races isn't the same as "I wrote code free of data races". > As for the “billion dollar mistake”, I understand the argument in the context of C or C++, but not in the context of Go. Even without a chance of UB. You're adding an implicit `Type | null` to every piece that uses nullable code without handling the null case. Each time you forget to do it, you cause either an UB or a null pointer error. And the place it manifests is different from where it's generated. Furthermore, listening to Tony Hoare's talk ( [this video] ), he mentions that trying to avoid the null, in the context of a language like Java or C#, that permits them is also causing people to waste time working around them. > The middle network layer will report the error to the user, I fix it, and move on, a $0 mistake. Sure, but that's not a $0 mistake. It's {time to fix * hourly rate}. Even if

> Even without a chance of UB. You're adding an implicit `Type | null` to every piece that uses nullable code without handling the null case. Each time you forget to do it, you cause either an UB or a null pointer error. And the place it manifests is different from where it's generated. > Furthermore, listening to Tony Hoare's talk ( [this video] ), he mentions that trying to avoid the null, in the context of a language like Java or C#, that permits them is also causing people to waste time working around them. It's an insignificant amount of time in my experience, which makes this irrelevant in practice. I’ve never had to "work around them", this is either a theoretical argument or just sloppy programming. > Sure, but that's not a $0 mistake. It's {time to fix * hourly rate}. Even if you're doing for OSS, you could have been spending that time doing something else. Oh, come on, you’re smart enough to understand that a few minutes every few months may not be exactly $0, but it’s such an insignificant value that it can be treated as $0. > Sure. But trading "ease of use" for "preventing errors" is not something I want in any language I use. Well, you can choose your trade offs, and you can let others choose theirs. So I guess you use Ada for everything? or maybe you use Coq to prove everything? Of course you don’t, you’re also making trade offs, just different ones. > Null/nil are about as good concepts today as silently skipping errors. You exaggerate again. I do not agree with tha