Should it be possible to update `submitter` with `updateSystemMetadata()`?

[rogerdahl]

https://releases.dataone.org/online/api-documentation-v2.0/apis/Types.html#Types.SystemMetadata.submitter

From @servilla

The "submitter" attribute in system metadata should be mutable through the updateSystemMetadata method call on the MN so that incorrect content on the MN can be modified to be consistent with what exists on the CN. This function is currently and quietly disabled through MemberNodeClient_2_0.updateSystemMetadata() on GMN v2.3.10.

[amoeba]

Interesting.

incorrect content on the MN

How does this happen, such that the CN's copy is "correct" while the MN's is not? IMO allowing submitter to be mutable limits the utility of the field.

[rogerdahl]

How does this happen, such that the CN's copy is "correct" while the MN's is not? IMO allowing submitter to be mutable limits the utility of the field.

Spec just says that the MN must set this field when the object is first created by a client. So where it should come from or set to is implementation specific. So "wrong" just means that someone's idea of how to best set this field changed, or a bug caused it to not be set to what was intended.

In the case of GMN, the field is normally created by generating a D1 subject string from the DN of the certificate the client used in the create call. But when GMN is used in a "slender node" configuration, the client becomes an integral part of the MN itself, so there are settings to have GMN start trusting the client to set values in the system metadata that we have specified have to set by the MN. One of those is the submitter, so original reason for this was probably an issue with how it was filled in by the slender node adapter (client).

[amoeba]

Ah, gotcha. Thanks for explaining. Curious what others think.