Joe Tsai

\cc @neild who has been interested in adding a "MutableFoo" accessor method which does exactly this.

The current solution is certainly not ideal, but I'm going to put this on hold for until there's a resolution on golang/go#19412.

By default we cannot enable this feature since the "SetXXX" method could conflict with an actual field named "SetXXX". This is not a hypothetical scenario, but a real problem. At...

The `protojson` package uses its own implementation of JSON since it needs to comply with a stricter version of JSON ([RFC 7493](https://tools.ietf.org/html/rfc7493)) than what `encoding/json` supports ([RFC 7159](https://tools.ietf.org/html/rfc7159)). The internal...

> The main reason that encoding/json is not used is because google/protobuf/duration.proto, int64(string) and oneof marshal are inconsistent with the standard library JSON and there is no good solution. That's...

Just wanted to note that it would have to be `*time.Duration` in order to preserve the fact that proto has a distinction between the zero duration and an unset duration.

This proposal is on hold until reflection work is finished (see #364) as generating native Go types will need to inter-operate with proto reflection in a reasonable way.

(Note for consideration in the future) We need to decide whether this function keeps a reference to a `protoregistry.Files`-like object or not. The semantic difference is whether `NewTypes` iterates through...

As an implementation note, I'm waffling whether to add a `FindExtensionByNumber` method to `protoregistry.Files` so that `NewTypes` can simply be a thin wrapper over the `Files`-like object that is provided....

This looks like a valid regression bug. Can you check with the [newer `protojson` package](https://pkg.go.dev/google.golang.org/protobuf/encoding/protojson) what the behavior is?