Oddbean

For when yall give up and realize C is the proper language to build in. https://www.vaughnnugent.com/resources/software/modules/noscrypt nostr:nevent1qvzqqqqqqypzpckv7l8jqspl8u4y54dn9rcduwlrs4v2040nxce0m2h0cunvrj8tqqsf8k2tmhsdyvp7v67y8a6k426kuyqn5sf690j4ujf58r30vyx3emqaea7gy

i wish i could spend 5 minutes walking you through the secp256k1 library used in bitcoin core and point out to you how many ways it is both excessively verbose, and confusing to follow, and then show you how people trained in using this language made the abomination called btcec, which is 1/10th as performant, as a result of their excessive verbosity and lack of brain

But why? It works and I have control over my CPU instructions if I want it :), because it's worked for 40+ years in the same way over and over again, and is available on every machine and developer and ABI layers, the same way it's always done. You can save your breath on zig I'm very aware of it's capabilities and I'm not against it, right tool for the right job, but for the reasons stated above, I think C is the largest hammer in the tool belt and should continue to be used when necessary.

Yeah, I simply avoid them. I used macros and tiny structs to track length. In noscrypt I replicated C# spans with inline functions and small structs, the same way literally every other language does because most of them are written in C anyway lol. To be fair though, if people didn't keep ignoring the language requirements, ub, and ignoring compiler manuals, storing the termination with 0 at the end of a buffer is arguably the most efficient way to handle that. Strings are just a sequence of memory, and literals are stored in data segments. Most efficient way to handle that imo. https://git.vaughnnugent.com/cgit/vnuge/noscrypt/.git/tree/src/nc-util.h#n96

> I used macros and tiny structs to track length. In noscrypt I replicated C# spans with inline functions and small structs, the same way literally every other language does because most of them are written in C anyway lol. the way you say that, C is mostly a language for people with too much time on their hands 😀 it's great if things like that just work with the standard types provided by the compiler/standard library, believe me, after 30 years of coding it doesn't sound like fun anymore to roll it yourself > storing the termination with 0 at the end of a buffer is arguably the most efficient way to handle that nah, PASCAL strings (storing the length explicitly) are slightly more efficient in most cases, as there's no need to iterate over the string again and again to get the length. But their biggest problem isn't efficiency, it's that a character (\0) is special-cased. This can often be exploited in various ways in protocol handling. (i say this as someone that has found tons of remote exploitable bugs in C code, including libminiupnp 😱 )

length prefixes is how i do all of my string and other array encoding in my binary data protocols Wurth was right about this, and notably, pretty sure that if you asked Thompson he would say that null termination was a false economy