Definitions of NDEFMessageInit and NDEFRecordInit violates Web IDL spec

[peria]

Two IDL dictionaries NDEFMessageInit and NDEFRecordInit are defined as following;

dictionary   {
  required sequence<NDEFRecordInit> records;
};
dictionary NDEFRecordInit {
  ...(snip)...
  NDEFRecordDataSource data;
};
typedef (DOMString or BufferSource or NDEFMessageInit) NDEFRecordDataSource;

This reference cycle does not meet the description "The type of a dictionary member must not include the dictionary it appears on." in Web IDL spec

[reillyeon]

I am curious about the reasoning behind the WebIDL specification intentionally banning recursive types. As a workaround we could define NDEFRecordDataSource as follows which should only trivially change the code developers need to write when creating embedded records,

typedef (DOMString or BufferSource or NDEFMessage) NDEFRecordDataSource;

[zolkis]

Right, good catch. How are we supposed to define recursive structures with Web IDL then?
Use interface with 3 constructors instead of dictionary for NDEFRecord?

[peria]

I think this issue is not urgent nor serious while most browsers are working well. :)
And I think it is a practical option to update Web IDL spec not to ban the cyclic references in dictionaries.

[rakuco]

This ended up being filed again as #620 after it bit us in the Chromium implementation :/

[reillyeon]

Shall we mark this as a duplicate?

[rakuco]

Yeah, I think that makes sense, as all the recent discussion's tracked there.