Skip to content
This repository was archived by the owner on Nov 21, 2026. It is now read-only.

Benchmarks: graph tables carry LDBC's "not an LDBC Benchmark Result" wording - #493

Merged
tae898 merged 1 commit into
mainfrom
benchmarks-ldbc-fair-use-wording
Oct 23, 2026
Merged

tae898 merged 1 commit into
mainfrom
benchmarks-ldbc-fair-use-wording

Conversation

@tae898

@tae898 tae898 commented Oct 23, 2026

Copy link
Copy Markdown

What

Every generated table built from LDBC data (l2 graph OLTP, l2olap graph OLAP, and restart for its graph rows) now opens with two registered sentences in export_web.py:

  • LDBC_NOTICE: "These are not LDBC Benchmark Results." plus the CC BY 4.0 credit to the Linked Data Benchmark Council, "we ran it ourselves, it was not audited by an LDBC-approved auditor, and LDBC has not endorsed it".
  • a per-table deviations sentence ("How this differs from the LDBC standard: ..."): which data and scale factors (persons-and-KNOWS projection at SF1 and SF10; the full SF1 network for the OLAP table), which queries are ours and which are LSQB's, no LDBC driver, no ACID or recovery test, no price/performance figure, our own Python harness, embedded and containerised engines under one cpuset and memory cap.

page_check._check_ldbc_notice is a new gate: a table in LDBC_DERIVED_TABLES (or whose dataset or size label names LDBC) that lacks the exact disclaimer or the deviations sentence fails the publish. PAGE-SPEC.md section 7 and its gate table describe the rule. Because the sentences are registered in OCT_PROSE, the existing condition gate also checks them for unpinned numbers (CC BY 4.0 is added to CONDITION_EXEMPT).

Why

LDBC's fair-use policy (https://ldbcouncil.org/benchmarks/fair-use-policies) lets only a result audited by an LDBC-approved auditor be called an LDBC Benchmark Result. Unaudited or derived work must say it is not one (for example "These are not LDBC Benchmark Results"), give CC BY 4.0 attribution, and say briefly how it differs from the standard. Implying LDBC endorsement is treated as trademark infringement. The positioning report found that none of this wording existed anywhere under benchmarks/, and the late-November paper and the DBBench page both show these tables.

Checked locally

uv run --no-project --with numpy --with pandas --with pytest python -m pytest on test_page_check_2611.py (three new tests: wording present on the fixture payload and free of unpinned numbers, removed wording caught, reworded disclaimer caught), test_export_client_twin.py, test_gen_engines_doc.py, test_publish_2611_e2e.py, test_refresh_web_page_2611.py and test_stage_generator_2611.py: all pass.

Not covered

The humemai/dbbench repo (docs only; no LDBC mention) and humem.ai/src/lib/dbbench/lanes.ts (no LDBC mention) are unchanged. Figures and the paper PDF generators were not touched; they need the same sentence in their captions when they are built from these tables.

🤖 Generated with Claude Code

…wording and their deviations, held by page_check

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@tae898
tae898 merged commit 9ec7002 into main Oct 23, 2026
4 checks passed
@tae898
tae898 deleted the benchmarks-ldbc-fair-use-wording branch October 23, 2026 17:38
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant