ZJX 1.22 NULLSPOOL: The Optimization We Refused to Ship
Study how ZJX evaluated an optimization that preserved archive correctness but failed end-to-end wall-time and CPU qualification, keeping the unshipped experiment out of production.
Study how ZJX evaluated an optimization that preserved archive correctness but failed end-to-end wall-time and CPU qualification, keeping the unshipped experiment out of production.
- Edition
- Qualification record 1.0
- Published
- Aug 31, 2026
- Preview updated
- Aug 30, 2026
- Reading time
- 11 minutes
- Full edition
- 3,400 words
- Status
- historical
Benchmark / Experiment
- Research question or engineering problem
- Did the NULLSPOOL candidate qualify for promotion as ZJX 1.22?
- Principal finding
- The candidate preserved archive correctness but failed end-to-end wall-time and CPU qualification, so it was not promoted or shipped.
- Evidence type
- Retained frozen-candidate qualification campaign and release decision.
- Method summary
- Evaluate the frozen candidate against baseline, keep invalid campaigns separate, verify archive correctness and release validation, and apply the fixed promotion gate.
- Scope
- Public lineage and disposition only. Exact measurements, mechanism analysis, and engineering lessons remain in the paid qualification record.
- Limitations
- The public preview does not publish campaign measurements or unreleased implementation mechanics.
- The disposition belongs to the bounded NULLSPOOL candidate and does not characterize later ZJX releases or every workload.
- Public source or reproduction note
- Current ZJX release line
- Published
- 2026-08-31
- Last verified
- 2026-08-30
- Status
- historical
Who this is for
- Compression and storage engineers reviewing negative performance results.
- Technical leads designing evidence-backed promotion and rollback gates.
- Builders who need to distinguish a successful proxy metric from a successful system change.
Not for
- Readers seeking unreleased ZJX source, private corpora, or a general benchmark of later ZJX versions.
Detailed contents
- 01
The NULLSPOOL hypothesis
Why the candidate looked promising before end-to-end qualification.
- 02
The targeted measurement boundary
What the proxy metric could and could not prove.
- 03
Archive correctness
Why exact bytes and restoration were necessary but insufficient.
- 04
Invalid campaign retention
How environmental invalidity remained separate from performance evidence.
- 05
The valid campaign
The end-to-end result that decided promotion.
- 06
Frozen-candidate validation
Why later repairs could not rewrite the tested candidate.
- 07
Failure analysis
Why removing targeted work did not produce a qualifying system improvement.
- 08
Durable release doctrine
The engineering rules retained after the code was rejected.
Named artifacts
Qualification summary
A bounded comparison of proxy, correctness, performance, and release-validation gates.
Campaign-validity ledger
A separation of invalid environmental evidence from the qualifying campaign.
Frozen-candidate boundary
A record of which repairs belonged to later code rather than the evaluated candidate.
Negative-result checklist
A reusable review surface for deciding whether an optimization earned promotion.
The missing version is part of the lineage
ZJX 1.22 was not accidentally skipped. NULLSPOOL occupied that research slot, preserved archive correctness, but failed end-to-end wall-time and CPU qualification. It was rejected, and ZJX 1.21 STREAMWALL therefore remained the current release line.
The public record should preserve that disposition because a release sequence is less credible when failed candidates simply disappear. The deeper campaign, measurements, and lessons remain part of the paid qualification record rather than the public index.
This boundary keeps the claim inspectable without publishing the full value of the experiment. The public can see that the gate said no; the full edition explains exactly why. It does not publish the campaign figures or diagnosis.
Public lineage boundary
- ZJX 1.21
- STREAMWALL qualified and remained the current release line.
- ZJX 1.22
- NULLSPOOL was evaluated, rejected, and not shipped.
- Public proof
- Disposition and release continuity remain visible.
- Paid record
- Campaign detail, measurements, failure analysis, and engineering lessons require an unlock.
Evidence and method
measured: Retained frozen-candidate qualification evidence separates archive correctness, campaign validity, end-to-end performance, and release validation before applying the promotion gate.
Limitations
- The public preview does not disclose the exact campaign measurements or unreleased implementation mechanics.
- The result belongs to the bounded NULLSPOOL candidate and does not characterize later ZJX releases or every workload.
Access and updates
Purchase includes lifetime read access to this edition, email-based recovery, and revisions published to the same edition.
Public companion: Review the current ZJX release line