RFC process?

[Mirek Rusin]

Is there any RFC process for json-rpc?

Any draft work for next iteration of the protocol open for contributions maybe?

We're using json-rpc extensively in production for many years with great success.

It would be great to get feedback on some extensions we've made:

1. error.code to be number | string instead of just number.

2. support for async generators - where larger array results don't suffer from head-of-line blocking, individual items are yielded and are interleaved with other messages.

[Matt (MPCM)]

No RFC process as such. Since the spec never went through that type of process originally and it was not really envisioned that an RFC process would be used to push changes forward. But that extensions to the specification should be popularized by users/community first. If they became popular enough and added enough value to the broader usage of the json-rpc concept, that a future version of the spec would be created (vs just spec + extension utilized by a few).

If you have behavior you deviated with and found valuable, posting here with details to share that would be great. Especially #2, though I am also curious about #1 also. :)

[Brad Jones]

I would +1 some method for collaboration beyond the discuss group. I also use JSON:API extensively and while it does not follow the IETF RFC process, the spec has evolved (with a recent 2.0 tagged) well with PRs against a codebase. See <https://github.com/json-api/json-api>.

There is also the issue of "truck factor" for those with control of the domain where the spec is published; people come and go and "things happen" and since JSON-RPC is widely used and treated as a spec to be respected, it's worth thinking about governance.

I don't think anything needs to be over-engineered, but these are the types of questions that impact the health of Internet infrastructure overall. I think this is the kind of thing that traditional RFCs mitigate against, of course that's at the expense of a lot of bureaucracy. I think it's easy enough to strike a middle-ground.

Brad