Even-numbered destination_port constraint

[knneth]

How does one best describe an IS-05 constraint that the RTP destination port number must be an even number (for a "good" RTCP-/IPMX-sender)?

The problem

• We initially constrained IPMX senders to the enum [5004, validated-but-configured-outside-of-NMOS-port], but got pushback from customers that more range is needed.
• Listing every conceivable port number in an enum isn't a feasible option.
• The pattern keyword is only valid for strings.
• If we reject odd-numbered ports between a minimum and maximum with an error message, then the NMOS testing suite will roast us.
• If we allow odd-numbered ports, then we're encouraging potential receiver incompatibilities, and the IPMX testing suite will roast us.

Potential solution
JSON schema defines a multipleOf keyword that seems fitting, but alas, this isn't currently allowed by constraint-schema.json AFAICT.

[garethsb]

Although I agree it would be nice to have a way to express this constraint, as you say, it's not realistic with the current API. However, the design of /constraints was intended to cover a useful majority not all eventualities, and so the spec says that rejecting a PATCH request that meets the constraints is perfectly valid:

A 500 response may be returned if there is some failure which was not caused by a violation of schema or constraints.

If the Testing Tool doesn't accommodate that, it's a flaw in the tool.