Yaacov Rydzinski

Note that we currently avoid direct validation that the implementations declared by the union do not conflict, i.e., the following will yield a validation error for the empty `SomeUnion` error...

Any independent rule would have to note that the following is valid: ```graphql union SomeUnion implements SomeInterface & AnotherInterface interface SomeInterface { commonField: String } interface AnotherInterface { commonField: String!...

My first pass of the algorithm for implementing the independent validation rule would be something like: 1. let `fieldDefsMap` be a map of field names to lists of field definitions...

Two additional thoughts: 1. I realized that objects and interfaces could also have a separate rule of this kind, but I don't believe they do, so I will pass. 2....

The same name fields are of different but overlapping types, so I do not think they collide except when trying to automatically assign fields to the Union. Similar to the...

> Well, it must be of FriendlyWithoutGreeting, because that's narrower. I guess this is where I messed up, it's actually of type "Friendly & FriendlyWithoutGreeting," so everything should be ok?...

Thank you so much @rivantsov @benjie for your feedback. I have tried to throw additional edge cases at this, and have failed to create a breaking change problem, although there...

Thanks for that awesome point! I guess how we want introspection to work is a good discussion for the working group. I thought giving unions fields would be across the...

I think we should consider being more explicit to start, and require matching depth of actual type in schema.

This comment surprises me. If this has been mooted several times... presumably we do not expect this to change?... and there is only one case of serial execution by spec......