How soon will __id come to the specification?

[calebmer]

I heard from @josephsavona that @leebyron had a proposal to add a global __id field to the specification similar to Relay’s Node id field. See here: facebook/relay#1061 (comment)

What's the status on such a field getting added to the specification and how can we speed that process up?

[josephsavona]

The larger context is that a small number of Relay users have designed their schema such that the id field is a type-specific identifier (e.g. the database primary key). For a cache, however, it is substantially simpler to use global ids for object identification, and having a standardized "id" field name as part of the spec would help developers to design a schema that could be used with multiple clients (without forcing every client to deal with customizable id names).

[calebmer]

Yep, I completely understand all the benefits and as I’ve been looking into Relay alternatives a global identification spec for the entire ecosystem seems like a great idea 😊. So I’m really excited for an __id specification for two reasons:

1. I am one such Relay user who has naming clashes with id, __id is an exciting solution.
2. I am a GraphQL user interested in building Relay alternatives and requesting a “blessed” specification from the GraphQL team endorsing __id 😊

[calebmer]

Last I heard here facebook/relay#1232 (comment), you (@josephsavona) and @leebyron said that an __id addition to the language is being considered. What are some factors under consideration and how can we make the specification definite?

[leebyron]

Yeah, we're working on this and I'll have an RFC pull request set up shortly where we can talk about it in more detail.

[calebmer]

🎉

[calebmer]

What’s the status on this @leebyron?

[Naoto-Ida]

This change would be amazing for our production GraphQL API. Can't wait to make it Relay compatible! Any updates would be greatly appreciated.

[leebyron]

Going to close this issue since it's a question and not a tracking RFC.

The concrete answer to the question is that __id will need to go through the RFC process, and that hasn't begun yet.

I definitely would like to address it before the April-2017 edition of the spec.

[stubailo]

@leebyron are you looking for someone to write up the __id RFC? I could take a pass at it.

[leebyron]

That would be great, @stubailo! Happy to collaborate on this one. We've spoken about some of the open concerns before, so perhaps you could start with the document we started a while back.

[borekb]

For anyone reading this issue, the RFC PR was #232.

My personal summary (please read the full discussion below the RFC yourself!) is that the __id field is not really a meta field in the same way that e.g. __typename is, for example, __id would be defined by users, it doesn't return a meaningful value for all types, etc. That's why it didn't get into the spec in that specific proposal.