msporny
commented
5 years ago Was any thought given to the fact this new protocol cannot be expressed using protocol buffers (or as natural objects/structs in most programming languages) because of the use of the non-standard "@" character?
Yes, a few:
- None of the current set of implementers are targeting protocol buffers as a target syntax -- it just has not come up over the past 6+ years as a target syntax.
- It is possible to serialize/deserialize to protocol buffers quite easily as the only
@translation that's necessary is for@context, so the rule is fairly simple and limited to that property, IIRC. - CBOR is a more likely target serialization, there is more interest in going to/from that than there seems to be for protocol buffers.
So, fundamentally, there is nothing serious that prevents serialization/deserialization to protobuf and we'd welcome a serialization profile that supports to/from protobuf. Are you interested in writing that spec or README?
I think this pretty much guarantees the fragmentation of the protocol with folks having to substitute
@contextwith things like_context,_at_context, etc.
To be clear, the spec is a data model and suggested serializations... it's not a protocol. To prevent fragmentation wrt. the @ symbol it seems like the only thing that would be required is a single serialization/deserialization rule. We're happy to write that up in the Implementation Guidance if you want to provide an implementation that serializes/deserializes to/from protobuf?
(The npm module https://www.npmjs.com/package/protobufjs alone has more than 1.7MM weekly downloads, and I would have thought this was the most likely audience for this kind of technology).
It doesn't seem to be the "mose likely" audience, given that the 300+ participants in the work to date have not identified protobuf as a target syntax. That said, if there is interested, we're more than happy to figure out a way of publishing interop specs for protobuf... given that the issue you've identified is the only one, I can't imagine it would be that difficult.
Are there other issues that you feel would be non-starters for a protobuf serialization?
burnburn
Was any thought given to the fact this new protocol cannot be expressed using protocol buffers (or as natural objects/structs in most programming languages) because of the use of the non-standard "@" character?
I think this pretty much guarantees the fragmentation of the protocol with folks having to substitute
@contextwith things like_context,_at_context, etc.(Reference: https://developers.google.com/protocol-buffers/docs/reference/proto3-spec#letters_and_digits)
(The npm module https://www.npmjs.com/package/protobufjs alone has more than 1.7MM weekly downloads, and I would have thought this was the most likely audience for this kind of technology).