Publication is part of the evidence chain
A research note may be factually careful on its first day and misleading six months later if its source vanishes, a major correction is made without a version marker, or a newer article quietly replaces an earlier claim. A professional publication therefore needs a record of what was asserted, what supported it, and what changed. A Git commit identifies a file revision; it does not establish that the revised assertion is true.
For each publication, keep a title, scope statement, date, source references, license or reuse conditions, author or editor attribution, version identifier and correction path. Treat editorial commentary and empirical findings as different types of output. A report of collected measurements needs its input and method; a theoretical essay must not present a speculative effect as measured.
Three layers of preservation
Original evidence: original documents, records or datasets with appropriate access and legal controls. For historical work this may mean archival references and page or folio identifiers rather than reposting a copyrighted scan.
Working interpretation: transcription, normalized dataset, translation, software output, note, and analytical reasoning. Document transformations rather than calling all derived objects “sources.”
Published artifact: reviewed article, public download or release package. Maintain an identifiable version and digest manifest where practical. Archiving a generated page preserves what was published, not the truth of its content.
How a correction should work
Record the affected URL and passage; describe the old claim and the proposed correction; identify the new source or test; mark whether the change is factual, typographical, methodological or a revised inference; and preserve the previous version where feasible. Critical errors should be visible at the top of the current work, not buried in an obscure changelog.
Corrections are not necessarily evidence of poor research. Refusing to separate facts from hypotheses, however, makes a correction nearly impossible to evaluate.
Rights and sensitive material
Public research archives can contain confidential material, regulated information and copyrighted work. The presence of a file in an investigation does not imply permission to redistribute it. Publish sanitized or synthetic fixtures where an educational example would otherwise expose protected records. Preserve real evidence under an authorized workflow, not in a public GitHub repository.
Current implementation
Benandanti.org publishes Markdown-derived static pages through a Git-managed workflow and includes a generated file-level SHA-256 manifest. The archive provenance page explains its limits. The Selected Publications index links to published web resources and labels research protocols as methods, not completed validation studies.
The publication workflow intentionally avoids a user-account database or on-site journal engine on shared hosting. Git history is useful for tracing editorial changes once the repository is actually created and maintained; this release does not claim an existing public commit history.
Source and revision record
Original editorial analysis and definitions. For background reference, consult the source catalog.