Changing an object to an interface isn't necessarily breaking

[eapache]

"the type changed" is almost always a breaking change, except when going from an object to an interface with exactly the same set of fields (and implemented interfaces). It would be nice if schema_comparator recognized this case and didn't flag it as breaking.

[jturkel]

Perhaps this should be a dangerous change since clients will get different results if they're selecting __typename for this object/interface?

[swalkinshaw]

Oh good point! Since the interface would return the resolved object type's name