<?xml version="1.0" encoding="UTF-8"?>
<rfc xmlns:xi="http://www.w3.org/2001/XInclude" category="info" docName="draft-watts-research-intelligence-archive-00" ipr="trust200902" submissionType="IETF" symRefs="true" tocInclude="true" tocDepth="3" version="3">
  <front>
    <title abbrev="Research Intelligence Archive">Evidence-Bounded Research Intelligence Archives</title>
    <seriesInfo name="Internet-Draft" value="draft-watts-research-intelligence-archive-00"/>
    <author fullname="Deonté O’dell Watts" initials="D. O." surname="Watts">
      <organization>Independent Researcher</organization>
      <address><email>deonte@goodshyt.fun</email><uri>https://orcid.org/0009-0005-8586-3650</uri></address>
    </author>
    <date year="2026" month="October" day="8"/>
    <area>General</area>
    <keyword>research provenance</keyword><keyword>evidence</keyword><keyword>reproducibility</keyword><keyword>archives</keyword>
    <abstract><t>This unsubmitted working draft defines an evidence-bounded archive profile for scientific and technical claims, sources, evidence tiers, falsifiers, tests, provenance, publication state, and integrity manifests. The profile is designed to prevent archival or mechanical integrity from being misread as empirical validity, novelty, peer review, standards approval, or publication acceptance. This file is an archival RFCXML working draft and has no IETF standing unless separately submitted.</t></abstract>
  </front>
  <middle>
    <section anchor="intro" numbered="true"><name>Introduction</name>
      <t>Research programs increasingly combine manuscripts, code, datasets, repository histories, protocol drafts, experimental records, figures, and publication metadata. A useful archive must preserve more than file storage: it must preserve which source supports which claim, what remains unknown, what would falsify a statement, and which gates remain before external release.</t>
      <t>This profile defines a compact interoperability model for those records. It is intentionally evidence-bounded. It does not define how scientific truth is established and does not replace venue-specific research-integrity, privacy, intellectual-property, security, or disclosure requirements.</t>
    </section>
    <section anchor="terms" numbered="true"><name>Terminology</name>
      <dl>
        <dt>Source Record</dt><dd>A record identifying an original or externally authoritative input, including a stable locator and, when available, a cryptographic digest.</dd>
        <dt>Claim Record</dt><dd>An atomic proposition associated with epistemic status, evidence tier, source references, falsifier, replication procedure, and release state.</dd>
        <dt>Evidence Tier</dt><dd>A local classification of evidentiary strength. This profile uses T1 for empirical/observational evidence, T2 for derivation, T3 for rigorous conjecture, and T4 for speculative material.</dd>
        <dt>State Collapse</dt><dd>An illegitimate promotion or merging of distinct epistemic or workflow states, such as requested to approved or executed to validated without required transition evidence.</dd>
        <dt>Mechanical Integrity</dt><dd>Evidence that bytes, builds, signatures, or hashes match expectations. Mechanical integrity is not empirical validation.</dd>
      </dl>
    </section>
    <section anchor="model" numbered="true"><name>Archive Data Model</name>
      <section numbered="true"><name>Source Registry</name><t>Each material source SHOULD receive a stable source identifier, original title, date or capture date, kind, authority description, trust label, archive treatment, locator, and digest where bytes are available. Multiple storage copies of the same intellectual record MUST NOT be counted as independent corroboration merely because their file identifiers differ.</t></section>
      <section numbered="true"><name>Claim Register</name><t>Each claim SHOULD be atomic enough to admit a falsifier. The record SHOULD contain a claim identifier, project identifier, status, evidence tier, source references, confidence rationale or bounded confidence field, falsifier, replication protocol, and publication-readiness state.</t></section>
      <section numbered="true"><name>Claim States</name><t>A profile MAY use ESTABLISHED, DERIVED, CONJECTURE, ANOMALY, and POTENTIALLY NOVEL. POTENTIALLY NOVEL is only a routing state for prior-art work and MUST NOT be interpreted as a novelty or priority certification.</t></section>
      <section numbered="true"><name>Evidence and Falsifier Registers</name><t>Evidence SHOULD be linked to claims without erasing provenance. Falsifiers SHOULD identify observations, proofs, prior art, or reproduction failures that would change the claim state. Negative and contradictory evidence MUST be retained.</t></section>
      <section numbered="true"><name>Publication State</name><t>Publication readiness SHOULD record the independent gates still required, for example empirical execution, independent reproduction, prior-art review, rightsholder review, security analysis, or venue acceptance. Repository visibility or a rendered manuscript MUST NOT be treated as venue acceptance.</t></section>
    </section>
    <section anchor="noncollapse" numbered="true"><name>Non-Collapse Requirement</name>
      <t>Distinct workflow and epistemic states MUST remain distinct unless an explicitly documented transition provides the required evidence. In particular, false and unknown are distinct; submitted and approved are distinct; authenticated and authorized are distinct; authorized and executed are distinct; executed and successful are distinct; successful and independently validated are distinct.</t>
      <t>For a state chain A to B to C, a consumer MUST NOT infer A directly implies C when the profile identifies evidence required for the omitted transition.</t>
    </section>
    <section anchor="provenance" numbered="true"><name>Provenance and Reconciliation</name>
      <t>An archive SHOULD preserve the original record, source identifier, revision or commit when available, capture date, and byte digest. Derived summaries MUST remain distinguishable from primary records. Source-derived context MUST remain separate from synthetic experiment state.</t>
      <t>When contradictions occur, the archive SHOULD preserve both records and record the discrepancy rather than silently harmonizing them. A reconciliation decision SHOULD identify the authority rule used.</t>
    </section>
    <section anchor="integrity" numbered="true"><name>Integrity and Reproducibility</name>
      <t>Implementations SHOULD compute cryptographic digests over archived files and SHOULD provide a manifest-verification procedure. Buildable manuscripts SHOULD include their source, bibliography, figures or stable figure references, and deterministic build instructions to the extent practical.</t>
      <t>A matching digest proves byte integrity only. A successful build proves build-level reproducibility only. Neither proves truth, authorship, causality, novelty, independent replication, peer review, publication acceptance, or standards adoption.</t>
    </section>
    <section anchor="interchange" numbered="true"><name>Interchange Formats</name>
      <t>JSON and CSV MAY be used as parallel machine-readable and human-reviewable tracker forms. BibLaTeX MAY carry bibliographic metadata. RFCXML v3 MAY be used for unsubmitted interoperability working drafts. A generated RFCXML file MUST NOT be described as an RFC or IETF-approved document unless it has separately achieved that status.</t>
      <t>Where multiple serializations exist, one serialization SHOULD be designated canonical for each logical register, or a deterministic conversion rule SHOULD be documented.</t>
    </section>
    <section anchor="privacy" numbered="true"><name>Privacy and Intellectual Property Considerations</name>
      <t>Archives SHOULD separate research records from credentials, secrets, unnecessary personal information, third-party intellectual property, and confidential business material. Credentials and private keys MUST NOT be included merely for reproducibility. Personal data SHOULD be minimized. Third-party authorship and licenses MUST be preserved.</t>
      <t>Publication of an archived source SHOULD occur only when the archive owner has the right to publish it or when an applicable license or legal basis permits redistribution. Internal archival possession does not create publication rights.</t>
    </section>
    <section anchor="security" numbered="true"><name>Security Considerations</name>
      <t>Archive tooling can become a high-value aggregation point. Implementations SHOULD use least-privilege access, separate public and confidential release sets, prevent accidental inclusion of credentials, and preserve an audit trail of source ingestion and mutation. Integrity manifests SHOULD themselves be versioned and protected against silent replacement.</t>
      <t>Archives MUST NOT silently delete contradictory or negative evidence in order to improve a narrative result. Automated transformations SHOULD operate on detached views when modification of canonical source records would compromise reproducibility.</t>
    </section>
    <section anchor="iana" numbered="true"><name>IANA Considerations</name><t>This document has no IANA actions.</t></section>
  </middle>
  <back>
    <references><name>Normative References</name>
      <reference anchor="RFC7991"><front><title>The xml2rfc Version 3 Vocabulary</title><author initials="P." surname="Hoffman"/><date year="2016"/></front><seriesInfo name="RFC" value="7991"/><seriesInfo name="DOI" value="10.17487/RFC7991"/></reference>
    </references>
  </back>
</rfc>
