property data types of vocab types are severely underspecified

[keighrim]

Because

While the MMIF spec is utterly delegating any responsibility to define data types for vocabulary at_types' property values to the vocabulary writer

mmif/specifications/index.md

Line 242 in fa34a55

The two required keys are `@type` and `properties`. As mentioned before, the `@type` key in JSON-LD is used to define the type of data structure. The `properties` dictionary typically contains the features defined for the annotation category as defined in the vocabularies at [CLAMS vocabulary ](vocabulary) or [LAPPS vocabulary](http://vocab.lappsgrid.org/). For example, for the *TimeFrame* annotation type the vocabulary includes the feature `frameType` as well as the inherited features `id`, `start` and `end`. Values should be as specified in the vocabulary, values typically are strings, identifiers and integers, or lists of strings, identifiers and integers.

Values should be as specified in the vocabulary, values typically are strings, identifiers and integers, or lists of strings, identifiers and integers.

, mmif-python is trying to define some (reasonable) upper bound in the complexity of the possible values.

However this only causes problems like clamsproject/mmif-python#252, and I'm becoming quite skeptical on maintaining the code in the python SDK that is not specified in the specification.

Done when

We decide to do either

1. update the spec with data types
2. get rid of type checker in the python SDK

and execute the decision.

Additional context

clamsproject/mmif-python#144