Again, I'm of the opinion that this vote is premature and needs to be halted.
Edmond had done extraordinary effort and proved that an async implementation is possible. And his solution may indeed be the route to go. But perhaps settle back and target this for 9.0 instead of 8.6, which will give whatever implementation is chosen freedom to make breaking changes should they be absolutely necessary. I'd also recommend an RFC to commit to having async in 9 in some form without committing to any specific implementation and set out the goals. I think passing that alone is significant. Furthermore, everyone remembers the PHP 6 debacle of committing to full unicode support and then discovering halfway into the process that such was a far bigger bear than realized. Edmond has explored the async problem space well enough I think that we can be sure that an implementation is possible. Rushing to commit in the first prototype though I feel is a mistake. Not that what I say has much weight - I'm a moron compared to most of you, but I try. On Tue, Nov 25, 2025 at 10:45 AM Edmond Dantes <[email protected]> wrote: > Hello > > > Anyway, it seems like we're coming to the conclusion that mixing sync > and async behavior is going to cause quite some problems and is something > that would not pass a vote if I had to guess. Even if you know what you're > doing YMMV. > > The main question is how much code needs to be changed. > The example I gave with `file_put_contents` is a typical source of errors. > It is easy to fix, but not always easy to detect. > > Maybe we can improve this situation. > This requires discussion. >
