Do we allow minItems:0 as an interim solution

[edeutsch]

The proposal is to switch how we say that the we have no attributes:

TRAPI 1.6 required empty array as the only way with minItems:0
ONLY: attributes: []

1. The proposal is to require a null value, and the array MUST NOT be empty: minItems:1
   OK: attributes: null
   OK: (absent)

2. A third compromise is to allow a transition grace period where no attributes may be expressed as the empty list from before OR the null value. We encourage everyone to move to null.
   OK: attributes: null
   OK: (absent)
   OK: attributes: []
   In TRAPI 2.1 we MAY fully switch to minItems:1

Vote for 1: tada

Vote for 2: rocket

[colleenXu]

Related to #506 (going to change this issue to "making attrib not required"), #514

[colleenXu]

Because of my STRONG objections to allowing and using null (which is a separate discussion in #514 (comment)), I'm thinking of the options this way instead:

1. 🎉 attributes: [] is NOT ALLOWED (minItems: 1)
   • If you want to express "no data", then you MUST use a different method (TBD on what is preferred/allowed)
   • This means previous behavior will be NOT VALID
2. 🚀 attributes: [] IS ALLOWED (minItems: 0)
   • Previous behavior will continue to be valid (for a transitional grace period)
   • STRONGLY ENCOURAGE everyone to use other methods to express "no data" (TBD on what is preferred/allowed)

[colleenXu]

Decision reached: keep minItems: 0 for these.