result.message SHALL be present constraint should be added to schema

[michaelcfanning]

Somehow we appear to have lost the schema constraint that result.message is required. The spec still makes this sensible condition clear.

[ghost]

@michaelcfanning Both the spec and the schema are correct. The spec says that either message must be present (if the log specifies its messages "inline"), or ruleMessageId must be present (if the log just provides a resource identifier together that locates the message string), or both.

Now even if the log file specifies a message indirectly via ruleMessageId, you still need message if the resource string has any replacement sequences {n} that need to be filled from message.arguments. But if the resource string has no replacement sequences, you don't need message at all.

So the schema can't require message.

[ghost]

This is actually a bug that crept in when we removed result.ruleMessageId. Before then, result.message could not be required, for the reason I stated above. But now, message is required because even a rule's message id is now stored in result.message.messageId, not in a separate result.ruleMessageId.

(The point being, this schema constraint wasn't "dropped"; it was never there. But now it can and should be there.)

@michaelcfanning FYI

[michaelcfanning]

Right. I your the reference entirely to a property that we removed.

Great. I can't help but think that our ability to add 'message' as a required property in the schema reflects that the design has improved slightly.

[ghost]

This is purely a schema change so no change draft is needed. I'll put it on the agenda to approve it.

[michaelcfanning]

The spec is correct and we should cover this explicitly in the editor's report. I don't see a need to require TC approval on this, however, we simply need to report out on the schema bug.

[ghost]

Agreed. I'd already removed it from the agenda and added it to the Editor's Report:

1. I made the following changes at editorial discretion:

   ...

   1. Issue #283: "result.message SHALL be present constraint should be added to schema"

      The spec text already says:

      A result object SHALL contain a property named message...

      It's just a bug that we neglected to mark the property required in the schema. The comment thread in the issue explains how this bug crept in.

      I added this change to the "schema changes" document so that Michael will fix it when he updates the schema with the changes for TC #28.

[michaelcfanning]

Updated the schema. Notification and result now both mark 'message' as a required property. @lgolding FYI

[ghost]

The schema correctly says that result.message is required.