annotations constraint behavior is surprising

[pbcornell]

The following examples illustrate annotations constraint behavior I find surprising (or not useful), and suggests a need to revisit the specification on this topic (e.g., what's the value of defining an optional annotation? how should an "open content" annotation be handled? etc.)

Defining optional annotations doesn't appear to add any value:

type::{name:pbc,type:symbol,annotations:[a,b,c]}
-------------------------------------------------------------------
            hello: true
      a::b::hello: true
   a::b::c::hello: true
a::b::c::b::hello: true
a::b::c::d::hello: true
   c::b::a::hello: true

Defining ordered, optional annotations doesn't appear to add any value (I think the impl is treating annotations that are out-of-order as open-content):

type::{name:pbc,type:symbol,annotations:ordered::[a,b,c]}
-------------------------------------------------------------------
            hello: true
      a::b::hello: true
   a::b::c::hello: true
a::b::c::b::hello: true
a::b::c::d::hello: true
   c::b::a::hello: true

The following examples illustrate different behavior when an annotation is repeated, or is open-content:

type::{name:pbc,type:symbol,annotations:required::[a,b,c]}
-------------------------------------------------------------------
            hello: false
      a::b::hello: false
   a::b::c::hello: true
a::b::c::b::hello: true   // repetition is allowed
a::b::c::c::hello: true   // repetition is allowed?
a::b::c::d::hello: true   // open content is allowed?
   c::b::a::hello: true

type::{name:pbc,type:symbol,annotations:required::ordered::[a,b,c]}
-------------------------------------------------------------------
            hello: false
      a::b::hello: false
   a::b::c::hello: true
a::b::c::b::hello: true   // repetition is allowed
a::b::c::c::hello: false  // repetition is denied?
a::b::c::d::hello: true   // open content is denied?
   c::b::a::hello: false

[popematt]

This surprising behavior will be left as-is in Ion Schema 1.0 (and so the implementation will not be "fixed"). The potential for this surprising behavior is eliminated in Ion Schema 2.0.