Joe Tsai

I should note that `ImportAs` would be best-effort, since we would presumably still prioritize correctness over cosmetics in the event that there is still a conflict over the chosen name.

As @neild said, protocol buffers do not support cyclic data structures. I'm not sure there's much we can do here. Detecting cycles is expensive, so we can't enable it by...

As the maintainer of the `encoding/json` package, I would argue that the depth used there is too low for protobuf, which tends to serialize much larger datasets. Secondly, the check...

\cc @neild

Removing `block-v2` label. We discussed this more today. * The various implementations of the v2 `proto.Message` today that exhibit this problem today panic for nil types. This makes it hard...

Most proposals to generate message fields as simply `M` instead of `*M` disregard protobuf correctness which requires that presence be preserved for these field. I appreciate that your proposal doesn't...

> Others in other forums have raised the point that regardless of the allocation concerns, the reflection-based decoding is dramatically less efficient than code-generated decoding. I’ve become convinced that we...

\cc @neild

Thanks for the report. The attached files do not appear to be `.tar.gz` files. Please advise.

Thanks, was able to generate SVGs for them. A bulk of the time in the new implementation is spent on "sizePointerSlow":  which implies either: 1) A bug in the...