Facebook hasn't used PHP for a very long time. It's been using Hack for over 12 years, which was initially based off PHP but it's deviated quite significantly now. It's now nearly entirely statically-typed, and has a bunch of features that PHP doesn't have. IO is fully asynchronous using async/await syntax (like C# or JavaScript), functions can be marked as pure, and a bunch of other things useful in large codebases.
It still uses the same request model as PHP, since that's genuinely useful for web requests. It's easier to avoid memory leaks if each request is entirely self-contained and there's no global state persisted between requests (other than APC and opcode caches)
It would be pretty wild if cloudflare or digital ocean opened themselves up to a massive lawsuit by doing this.
Do you have any evidence that cloud providers that don’t train their own models are violating their own contracts and risking their own good reputations by stealing data?
I hate to say this and this is because I fucking despise meta. But between DeepSeek and Meta, and trust they handle the training data correctly, I trust meta.
It still uses the same request model as PHP, since that's genuinely useful for web requests. It's easier to avoid memory leaks if each request is entirely self-contained and there's no global state persisted between requests (other than APC and opcode caches)