I haven't seen Defold before, but the message-passing architecture looks very interesting, so thanks for bringing it to my attention.

I have a couple of random questions, in case someone who knows the answers sees this:

1. Why protobuf and not Cap'n'Proto or Flatbuffers? Last I checked, both were better at zero-copy decoding, which might be a big deal for an engine with a message-passing architecture. Cap'n'Proto has had a lot of improvements for RPC use over the past few years.

2. Why HTTP/S rather than UDP? Both in terms of throughput and latency, UDP tends to come out ahead. Was it just not an issue?

3. Why WebP for compressed textures rather than KTX2, which offers BasisU supercompression? Doing so can get similar compression and greatly improve performance when decoding to GPU formats.

4. Any plans to support GLTF2? It has potential to replace custom engine formats, at least those that don't store data for easier GPU consumption, especially if used with meshopt. Unless the Defold format stores data in a GPU-ready format, it might be a good future option, and doing so would simplify asset pipelines a bit.