222 karma · joined April 14, 2020
Sure, data would need to be verified by a trusted party when it's first entered. A blockchain still protects against forgery after this though.
Compare this to any operation in Sublime Text/Merge and the difference is obvious.
The implication is also that more battery power is used for every action (not to mention vastly more memory), which matters for machines of any capability; but these are not latency-related concerns.
Particularly great is Zig's handling of both macros and generic types, the answer to both of which seems to be: just evaluate them at compile-time with regular functions, no special DSL or extra syntax. Andrew mentions in the video a big drawback of this system - implications for IDE complexity and performance. I imagine the performance side of this could be (maybe is?) mitigated by limiting recursion depth/loop counts for compile-time work.
I'm not particularly interested in taking on a language with manual memory management and the responsibilities it entails, but I would love to have access to Zig's compile-time capabilities, if it were available with some more memory safety.
And I think Xeon L3 cache tops out at about 40MB, whereas Threadripper & Epyc go up to 256MB.
> I often see many people show immediate solidarity with Amazon as soon as the ideas of anti-trust or monopoly are brought up. "If the government is allowed to break up a large corporation then where does it stop? Will they come after my small business next?"
Nowadays, 3-bit (TLC) and 4-bit (QLC) are much more common (not sure QLC even existed back then). So whilst each of these technologies has matured, the drive a consumer may consider today is more likely to be bigger, replacing their HDD entirely, and will probably be capable of fewer writes than an MLC drive from 5 years ago, iiuc.