| 16:59 | <Mathieu Hofman> | To clarify, when you say O(length), it's for the length of the string description, right? Not the length of the registry. I honestly doubt the length of the string is ever going to be meaningful in most normal cases. |
| 20:10 | <Mathieu Hofman> | While understanding the status quo is informative, my general concern is how much should that influence design decisions. I think this is a general problem for web dev. People build solutions based on observations of current performance behavior, which ultimately ends up validating the implementor's choice to not optimize certain cases. In this case, my understanding is that the semantics of a registered symbol are the ones that match the use case better, but there is some concern that it could cause some GC pressure in some (most?) implementations. Since I don't believe the GC issue is a intrinsic limitation, I think the solution shouldn't be designed around this constraint. |
| 21:35 | <kriskowal> | I’m not concerned about performance, but a registry stuffing attack. |