Re: [core] Comments on draft-ietf-core-yang-cbor-00

[Carsten Bormann <[e-mail]>]

Here is a nice example of a union that is actually a union in the
set-theoretic sense:

  typedef security-model {
    type union {
      type enumeration {
        enum v1  { value 1; }
        enum v2c { value 2; }
        enum usm { value 3; }
        enum tsm { value 4; }
      }
      type int32 {
        range "1..2147483647";
      }
    }
    reference
      "RFC 3411: An Architecture for Describing Simple Network
         Management Protocol (SNMP) Management Frameworks";
  }


(ietf-snmp-common.yang from https://github.com/YangModels/yang)

Note that there is no intention to actually allow a value of 1 to 4 for
the int32 arm.

This is an example of a general pattern that we see in data modeling
languages where the model both tries to define the generic set of values
(positive int32 here) and specific allocations within that set.
(CDDL has .within for this purpose.)

Here it would not be useful to tag the enumeration specially, as all
positive int32 values are intended to be allowed; it is just that four
of those values have a meaning defined in this specification.  If that
specification evolves, we might add more such values in the enumeration,
changing the tag for those with respect to implementations that
previously already have used that value without the specification having
specifically called it out.  Ouch.

(There are 219 occurrences of "type union" in that repo; most of these
are just DRY violations though.)

Grüße, Carsten