compliance: publish the EU draft PUE bands as a sourced data pack, not as code #78

Merged

dcharlot merged 1 commit from compliance/eu-scale-pack into main 2026-09-06 08:19:10 -04:00

Conversation 0 Commits 1 Files changed 4 +112

dcharlot commented 2026-09-04 05:52:39 -04:00

Owner

Copy link

BandingScale deliberately ships no canonical scale: which thresholds are correct for a metric is a policy question a library has no standing to settle. Baking one in would repeat the mistake the type was written to retire — is_compliant's pue < 1.5 && renewable_pct >= 50.0, two magic floats with no provenance, no version, and no way to say which instrument they encoded.

So the EU draft's PUE bands are published as data, in jouleclaw/scales/v1/, with their sources beside them. BandingScale already derives Deserialize, so a pack loads with serde_json::from_str — this adds no crate code at all, which is the point.

What the provenance file says plainly

• The instrument is a DRAFT — registered 2026-03-26, revised 2026-07-02, not adopted, not in the Official Journal, first labels expected 2027-08-15. A grade produced from it is not a regulatory determination.
• WUE is deliberately absent. The instrument sets the grade from PUE and WUE; only the PUE bands were read from it. Inventing WUE boundaries to complete the set would be exactly the failure this design prevents — a threshold with no provenance, presented as though it had one. A test asserts a composite built from this pack reports covers(&["pue","wue"]) == false, so a half-graded facility cannot read as graded.
• The instrument permits calculation and estimates, not only measurement, so a PUE fed to this scale may be self-declared. The scale bands whatever number it is given and says nothing about how that number was acquired.

The scale id is a version, never edited in place

Ratings cite their scale_id, and comparable_with refuses to rank across scales — which only works if an id means one fixed set of boundaries forever. A changed instrument is a new id, never an edit that would silently re-band grades already issued.

Tested, not merely committed

The pack is include_str!-ed, deserialized, validated and banded. The test pins that its last boundary is 1900 — this file's own example fixture uses 2000, and only one of the two is sourced.

52 jouleclaw-compliance tests pass (2 new).

🤖 Generated with Claude Code

`BandingScale` deliberately ships **no canonical scale**: which thresholds are correct for a metric is a policy question a library has no standing to settle. Baking one in would repeat the mistake the type was written to retire — `is_compliant`'s `pue < 1.5 && renewable_pct >= 50.0`, two magic floats with no provenance, no version, and no way to say which instrument they encoded. So the EU draft's PUE bands are published as **data**, in `jouleclaw/scales/v1/`, with their sources beside them. `BandingScale` already derives `Deserialize`, so a pack loads with `serde_json::from_str` — **this adds no crate code at all**, which is the point. ## What the provenance file says plainly - **The instrument is a DRAFT** — registered 2026-03-26, revised 2026-07-02, **not adopted**, not in the Official Journal, first labels expected 2027-08-15. A grade produced from it is *not* a regulatory determination. - **WUE is deliberately absent.** The instrument sets the grade from PUE *and* WUE; only the PUE bands were read from it. Inventing WUE boundaries to complete the set would be exactly the failure this design prevents — a threshold with no provenance, presented as though it had one. A test asserts a composite built from this pack reports `covers(&["pue","wue"]) == false`, so a half-graded facility cannot read as graded. - **The instrument permits calculation and estimates**, not only measurement, so a PUE fed to this scale may be self-declared. The scale bands whatever number it is given and says nothing about how that number was acquired. ## The scale id is a version, never edited in place Ratings cite their `scale_id`, and `comparable_with` refuses to rank across scales — which only works if an id means one fixed set of boundaries forever. A changed instrument is a **new id**, never an edit that would silently re-band grades already issued. ## Tested, not merely committed The pack is `include_str!`-ed, deserialized, validated and banded. The test pins that its last boundary is **1900** — this file's own example fixture uses 2000, and only one of the two is sourced. 52 jouleclaw-compliance tests pass (2 new). 🤖 Generated with [Claude Code](https://claude.com/claude-code)

dcharlot added 1 commit 2026-09-04 05:52:39 -04:00

compliance: publish the EU draft PUE bands as a sourced data pack, not as code ... e1a282dd49

`BandingScale` deliberately ships no canonical scale: which thresholds are
correct for a metric is a policy question a library has no standing to settle,
and baking one in would repeat the mistake the type was written to retire —
`is_compliant`'s `pue < 1.5 && renewable_pct >= 50.0`, two magic floats with no
provenance, no version, and no way to say which instrument they encoded.

So the EU draft's PUE bands are published as **data**, in `jouleclaw/scales/v1/`,
with their sources beside them. `BandingScale` already derives `Deserialize`, so
a pack loads with `serde_json::from_str` and this adds no crate code at all —
which is the point.

Each pack is exactly a `BandingScale`, accompanied by a provenance file
recording where the numbers came from, the status of that source, what the pack
omits, and the reuse terms. A pack with no provenance file is not publishable
here; a threshold whose origin is unstated is the thing the directory exists to
replace.

Three things the provenance says plainly:

- **The instrument is a DRAFT** — registered 2026-03-26, revised 2026-07-02, not
  adopted and not in the Official Journal, with first labels expected
  2027-08-15. A grade produced from it is not a regulatory determination.
- **WUE is deliberately absent.** The instrument sets the grade from PUE *and*
  WUE, and only the PUE bands were read from it. Inventing WUE boundaries to
  complete the set would be exactly the failure this design prevents: a
  threshold with no provenance, presented as though it had one. A test asserts
  that a composite built from this pack reports `covers(&["pue","wue"]) ==
  false`, so a half-graded facility cannot read as graded.
- **The instrument permits calculation and estimates**, not only measurement, so
  a PUE fed to this scale may be self-declared. The scale bands whatever number
  it is given and says nothing about how that number was acquired.

The scale id is a version and is never edited in place. Ratings cite their
`scale_id` and `comparable_with` refuses to rank across scales, which only works
if an id means one fixed set of boundaries forever — so a changed instrument is
a new id, never an edit that would silently re-band grades already issued.

The pack is tested, not merely committed: it is `include_str!`-ed, deserialized,
validated and banded, and the test pins that its last boundary is **1900** —
this file's own example fixture uses 2000, and only one of the two is sourced.

Verified: 52 jouleclaw-compliance tests pass (2 new).

Co-Authored-By: Claude Opus 5 <[email protected]>

dcharlot force-pushed compliance/eu-scale-pack from e1a282dd49 to fdc528e90c 2026-09-04 06:35:28 -04:00 Compare

dcharlot force-pushed compliance/eu-scale-pack from fdc528e90c to 4d9236fe1c 2026-09-06 06:24:50 -04:00 Compare

dcharlot merged commit dd3e43d802 into main 2026-09-06 08:19:10 -04:00

dcharlot referenced this pull request from a commit 2026-09-06 08:19:11 -04:00

compliance: publish the EU draft PUE bands as a sourced data pack, not as code (#78)

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels

No items

No labels

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

Transaction-Science/open-standards!78

No description provided.