Do not update child records using parent's E-tag

User Story: Jira Story

When we have an endpoint that updates a record in the db, it's sometimes desirable to update a child record as well.

Generally, to update a record, the caller must provide an E-tag, passed in the header If-Match that matches that of the record in the db.

However the parent and child have two different E-tags, and only the parent's E-tag is passed in a sole parameter in the header.

Therefore, it's not possible to pass in the child and parent E-tag cleanly.

Considered Alternatives

• Make a new endpoint for child updates so they can be updated separately with the correct E-tag.
• Pass a second E-tag in the body if the child is to be updated.
• Bubble up a child's updated_at value to the parent, so that the child and parent will have one E-tag.

Decision Outcome

We will make a new endpoint for child updates so they can be updated separately with the correct E-tag.

Currently this is just true for address and agent updates.

Pros and Cons of the Alternatives

Make a new endpoint for child updates

• + The mechanism for optimistic locking stays the same across all endpoints, so it's understandable for the Prime.
• + The updated_at value for parent and child record will correctly state the last time that record was updated.
• - More endpoints to create and maintain.

Pass a second E-tag in the body if the child is to be updated

• - The mechanism differs when you want to update a child, as the E-tag is passed in body instead of header. Makes the mechanism inconsistent, harder to reason about and harder to explain to Prime.
• + Fewer endpoints to maintain

Bubble up a child's updated_at value to the parent

• - Adds complexity because the child may have multiple parents and Prime would not realize that they have unwittingly updated unrelated records.
• - The mechanism differs from the norm and making exceptions for certain updates, will make it hard to be consistent across the codebase.
• + Fewer endpoints to maintain