Problem
The evaluation run-3 found two gaps in the Weaver semantic convention registry:
1. SCH-002: Ad-hoc attributes not in registry
Two attributes were invented by the instrumentation agent because they weren't in the registry:
commit_story.git.subcommand — records which git CLI subcommand is being traced (show, diff-tree, rev-list, log)
commit_story.commit.parent_count — number of parent commits (1 for regular, 2+ for merges)
Both are legitimate domain attributes that provide useful observability context.
2. CDQ-007: PII annotation on commit_story.commit.author
commit_story.commit.author stores the git author name, which is personal data per CDQ-007. Since git author names are public metadata on every commit and author attribution is core to the journal's purpose, the decision is to keep the attribute and add a note field documenting the PII acceptance.
Changes needed
In semconv/attributes.yaml:
- Add
commit_story.commit.parent_count (int) to the commit attributes group
- Add new
registry.commit_story.git attribute group with commit_story.git.subcommand (string)
- Add PII acceptance
note to commit_story.commit.author
Context
- These changes are already applied in the eval repo (commit-story-v2-eval) for run-4
- Upstream commit-story-v2 should be the source of truth for the registry
- Orbweaver issues #102 and #147 fixed the agent-side handling of schema extensions
Problem
The evaluation run-3 found two gaps in the Weaver semantic convention registry:
1. SCH-002: Ad-hoc attributes not in registry
Two attributes were invented by the instrumentation agent because they weren't in the registry:
commit_story.git.subcommand— records which git CLI subcommand is being traced (show, diff-tree, rev-list, log)commit_story.commit.parent_count— number of parent commits (1 for regular, 2+ for merges)Both are legitimate domain attributes that provide useful observability context.
2. CDQ-007: PII annotation on commit_story.commit.author
commit_story.commit.authorstores the git author name, which is personal data per CDQ-007. Since git author names are public metadata on every commit and author attribution is core to the journal's purpose, the decision is to keep the attribute and add anotefield documenting the PII acceptance.Changes needed
In
semconv/attributes.yaml:commit_story.commit.parent_count(int) to the commit attributes groupregistry.commit_story.gitattribute group withcommit_story.git.subcommand(string)notetocommit_story.commit.authorContext