Re: [VOTE] [RFC] Bound-Erased Generic Types

Le 15/06/2026 à 10:51, Rowan Tommins [IMSoP] a écrit :

On 15 June 2026 09:20:09 BST, Daniil Gentili <daniil.gentili@gmail.com> wrote:

I agree 100%, voters should take responsibility for their own actions, instead of asking to hold off the vote indefinitely, constantly requesting changes.

I think it's generally preferable for the main aim to be reaching consensus, and a vote confirming that consensus, rather than treating votes as something to be "won" and "lost". However, if there's genuine deadlock in a discussion, I guess bringing to a vote is a way to get a decision made.

The discussion thread on the internals list is already huge, so I guess at least a substantial proportion of the people who voted have already expressed themselves. And to what I've read on the thread, the decision were made on *this RFC*, not on the concept of Generics itself.

If this RFC is rejected, PHP won’t have generics (at least not in 2026/the first half of 2027 due to the cooling off period of RFCs of the same type).

That is, luckily, not what the policy says; it says "it will not be allowed to bring up a rejected proposal for another vote, unless ... the authors make substantial changes to the proposal". So an RFC building on Seifeddine's work but taking a different approach to enforcement could be brought forward at any time. Holding a vote is not, and should not be, a way to block alternatives or counter-proposals. Let's not get into the rhetoric of "this is your only chance for generics". If the vote is going ahead, people should vote on *this specific proposal*, with the specific tradeoffs it includes. It's great that we're making progress towards generics; now let's work together to get the best version of them we can.

One of my interpretations is that "complete generics" isn't possible per-se considering the current way the compiler and engine work. However, it's not impossible to implement "generic-like systems", and maybe a smaller RFC, for example about typed arrays/ArrayObject/Iterator, could already bring a lot of value, without having full generics yet.