eduKateSG · VOYNICH RESEARCH LIBRARY · TANGENTIAL VOYNICH VI
Tangential Voynich | Treat the Manuscript as a Filesystem
Physical pages are not necessarily logical neighbours. Paths are not storage addresses. Links can make one object appear in several places. What happens when an unread codex is forced to behave like organised storage?
← Tangential Voynich: The Wrong-System Protocol · Previous False World: Database · Voynich Research Library
A codex invites a powerful assumption.
Page 20 follows page 19.
The recto belongs before the verso.
A quire has an order.
Turn enough pages and eventually you have traversed the book.
That feels obvious because books are physical objects and reading is often linear.
But a filesystem immediately destabilises the assumption.
On a computer, the file you open as /research/voynich/test1/report.txt does not need to occupy one continuous stretch of physical storage. Its blocks may be scattered. Its logical parent may be a directory structure that exists only through metadata. Two paths may point to the same underlying object. A mounted filesystem may appear inside another directory tree while physically residing somewhere else. Deleted data may survive after its directory entry disappears. A damaged index can make intact data temporarily unreachable.
In other words:
Logical organisation and physical location are not the same thing.
That is the fifth false world.
The Voynich Manuscript is not a filesystem.
But filesystem thinking asks a dangerous and potentially productive question:
What if the surviving physical order of the manuscript is not identical to its logical organisation?
That question already matters in codicology.
The filesystem lens makes it much harder to ignore.
FALSE-WORLD CONTRACT
Voynich Is Not a Filesystem
- A quire is not a directory.
- A folio number is not a pathname.
- A page is not a file.
- A bifolium is not a disk block.
- A repeated label is not a symbolic link.
- A missing leaf is not a deleted file.
- A foldout is not a mounted volume.
- A scribal hand is not a user account.
- A physical discontinuity is not automatically fragmentation.
- A modern storage model does not establish medieval function, authorship, language or provenance.
Filesystem vocabulary is temporary. It survives only long enough to produce a measurable prediction.
Path is not provenance. Logical adjacency is not physical adjacency. Storage order is not reading order.
What the Filesystem Adds to Tangential Voynich
The database asked whether distant structures can share relationships.
The filesystem asks something more specific:
Can one logical structure survive despite physical scattering, reordering, renaming or partial loss?
This is a particularly strong wrong world because codices really can be rebound, lose leaves, acquire later foliation and preserve conjugate relationships that do not align neatly with the current reading sequence.
The filesystem lens therefore introduces several new questions:
- What is physical location versus logical membership?
- Can two non-adjacent pages be logical neighbours?
- Can one surviving element belong to several logical paths?
- Can a missing element be inferred because a path terminates unexpectedly?
- Can metadata preserve identity when content is ambiguous?
- Can a later index impose a new traversal order without changing underlying content?
- Can one logical collection be fragmented across several physical regions?
These questions are not translations.
They are reconstruction problems.
Physical Blocks Versus Logical Files
A file can be logically continuous while physically fragmented.
The user sees one document.
The storage device may see pieces.
This distinction maps unusually well onto one of the hardest Voynich problems: present page sequence versus production structure.
Suppose two folios are distant in the current binding but share:
- the same proposed hand;
- the same visual morphology;
- the same distributional cluster;
- the same boundary regime;
- recurrent rare forms;
- related damage or material characteristics.
That does not prove they were adjacent.
But filesystem thinking says physical distance should not automatically weaken a logical relation.
Conversely, two adjacent folios may be physically near while structurally unrelated because later binding placed them together.
Current neighbourhood should be tested, not assumed.
This connects directly to The Original-Order Problem.
Directories: Hierarchy May Be Imposed Rather Than Visible
A directory tree organises files into a hierarchy.
The files themselves do not need to contain their full path.
This creates a useful challenge to section naming.
Researchers often divide Voynich into sections because groups of pages share visual characteristics. That can be useful descriptively. The danger comes when a visual directory is mistaken for semantic identity.
Filesystem thinking asks whether several hierarchies could organise the same pages.
- By quire.
- By bifolium.
- By hand.
- By Currier regime.
- By morphology.
- By geometry.
- By distributional cluster.
- By interface type.
No one directory tree must be canonical.
This is precisely what the frozen Test 1b result suggests: the manuscript resists one universal partition and is better represented through overlapping memberships.
The neutral return is therefore:
A page may belong to several legitimate organisational hierarchies simultaneously.
The filesystem gives us permission to stop demanding one folder tree for the whole object.
Paths: One Object Can Be Reached Through Different Routes
A file has a path because the filesystem provides a route through directories.
That route is not the same thing as the file’s physical storage location.
This gives us a powerful model for Voynich retrieval and interpretation.
The same page can be reached intellectually through several routes:
- material object → quire → bifolium → folio;
- scribe → hand population → page;
- visual morphology → component family → page;
- text cluster → recurrent form → page;
- interface class → label behaviour → page.
If several independent paths converge on the same small set of pages, that convergence is structurally informative.
The database article called this candidate-key convergence.
The filesystem adds a navigational interpretation:
Robust structural objects should remain discoverable through more than one reasonable analytical path.
If a claimed population disappears when one route changes, it may be an artefact of the directory we chose.
Hard Links: One Underlying Object, More Than One Location
A hard link allows multiple directory entries to refer to the same underlying file object.
This generates one of the stranger but more useful Voynich questions.
Can one underlying structural entity appear in more than one apparent context without being duplicated?
For instance, a recurring label family may appear beside different visual forms. A repeated text family may occur in both prose and diagram interfaces. A visual component may recur inside several larger drawings.
Instead of assuming these are independent occurrences, hard-link thinking asks whether they could be multiple references to one underlying structural class.
The test must be stringent.
- Does the candidate preserve a stable neighbourhood role?
- Does it connect the same higher-order population?
- Does it retain similar positional constraints?
- Does the relation survive page type and hand differences?
If not, “same object” is too strong.
If yes, the neutral return is a recurring structure with cross-context identity.
Symbolic Links: A Marker Can Point Somewhere Else
A symbolic link is not the target object. It contains a reference to a path.
This is an especially useful model for short labels and navigational marks.
A tiny inscription may not contain the full semantic payload of the nearby image. It may identify, classify, cross-reference or redirect the trained reader toward knowledge located elsewhere—in the manuscript, in memory, or in an external tradition.
But the metaphor must generate predictions.
- Reference-like forms should recur across separated contexts.
- They should associate with stable target classes more than chance.
- They may be short relative to surrounding prose.
- Their local neighbour grammar may differ from running text.
- The same form may connect distant pages or interface classes.
If a short form has no stable relational target, the symbolic-link analogy fails.
If stable cross-context association survives, the neutral result becomes a candidate reference-like form.
This strengthens the database article’s foreign-key question from a different world.
Broken Links: Missing Material Can Reveal Former Structure
A symbolic link can survive after its target disappears.
The link becomes broken.
This gives filesystem thinking a powerful relationship to missing leaves.
Suppose a recurring cross-page structure strongly predicts a counterpart. Most instances have one. A small number do not. If those failures disproportionately occur near damaged, missing or uncertain codicological regions, the missing counterpart becomes reconstructively interesting.
The database called this an orphan relation.
The filesystem calls it a broken link.
The neutral measurement is stronger than either metaphor:
Expected relational counterparts may help identify loss, displacement or false reconstruction.
This can be tested directly against the Missing Leaves and Quire Reconstruction lane.
Fragmentation: One Logical Unit May Be Physically Scattered
Filesystems may store one file in non-contiguous blocks.
From the user’s point of view, the file is continuous.
From the storage system’s point of view, it is fragmented.
This is an excellent adversarial lens for the assumption that a functional manuscript section must occupy one continuous run of folios.
Suppose a structural page family appears in several separated codex regions but preserves a distinctive combination of hand, text behaviour, geometry and interface.
Three explanations might compete:
- the same logical programme is physically fragmented;
- the same template was reused in multiple independent programmes;
- the apparent similarity is generic and carries no logical identity.
The filesystem metaphor does not choose.
It demands a fragmentation test.
Do separated instances resemble one another more strongly than matched controls?
Do they preserve internal ordering?
Do they share rare features?
Do codicological clues support former proximity?
If not, fragmentation dies.
If yes, physical separation should no longer be treated as decisive evidence of logical separation.
Allocation Tables: Can We Reconstruct Continuity From Metadata?
Some filesystems track which storage blocks belong to which files through allocation metadata.
A codex has its own forms of structural metadata.
- conjugate leaves;
- quire signatures;
- foliation;
- stitching and sewing structure;
- parchment-side patterns;
- damage correspondence;
- contact transfer;
- hand transitions.
None of these is an allocation table.
But filesystem thinking asks whether these metadata channels can be combined to reconstruct probable continuity independent of current folio order.
A useful reconstruction should require convergence:
- physical evidence supports the relation;
- textual transitions do not contradict it;
- visual structure is compatible;
- hand sequence is plausible;
- the proposed order reduces unexplained breaks elsewhere.
This turns “these pages feel like they belong together” into a multi-channel reconstruction problem.
Inodes: Identity Can Be Separate From Name
In many filesystems, the underlying file object has metadata identity separate from the human-readable filename.
This is a valuable conceptual tool for Voynich.
A folio number is a later human-facing label.
The physical leaf has identity independent of that number.
Likewise, a modern section name such as “herbal” or “recipes” is a label applied by observers. The underlying page has structural properties independent of the name.
Filesystem thinking therefore reinforces an important rule:
Identifier, display name and inferred function should be stored as separate analytical fields.
This seems mundane.
It is not.
Once a page is named “pharmaceutical,” every later observation risks being interpreted through that folder name.
Neutral identity must precede semantic naming.
Mount Points: One Logical World Can Appear Inside Another
A filesystem can be mounted inside another directory tree.
To the user, it looks like one path hierarchy.
Underneath, distinct storage systems meet at a mount point.
This generates a provocative model for composite manuscript construction.
What if one codex combines originally distinct production units that were later made to appear continuous?
Possible mount-point-like boundaries would need multiple independent signatures:
- hand change;
- quire boundary;
- material change;
- visual programme shift;
- distributional discontinuity;
- layout transition;
- numbering or binding evidence.
One change alone is insufficient.
Several aligned changes may indicate a real production boundary.
After metaphor removal, the claim becomes:
Some apparent continuities may join originally distinct production units.
That is a codicological question, not a computing claim.
Permissions: Visibility Does Not Equal Intended Access
Filesystems can restrict who may read, write or execute an object.
The manuscript obviously contains no Unix permission bits.
But the wrong world forces a receiver question.
Was every layer intended for every reader?
A professional manuscript may contain shorthand, workshop marks, navigation cues, owner annotations or expert conventions that are visible to everyone but usable only by trained receivers.
We can test interface populations rather than assume universal readability.
- Do some marks occur only in one hand?
- Do some label forms appear only beside particular visual classes?
- Do some symbols cluster at production boundaries?
- Do later annotations use a clearly different system?
The neutral return is not “permission.”
It is receiver- or production-specific interface restriction.
Hidden Files: What If Important Structure Is Not Visually Prominent?
Filesystems can contain hidden files—ordinary objects deliberately omitted from routine display.
The Voynich analogue is not literal hiding.
It is normalcy blindness.
Researchers naturally focus on spectacular elements: strange plants, circular diagrams, human figures, the Rosettes foldout.
But the most informative structure may be visually boring:
- line length;
- gap width;
- margin behaviour;
- repeated minor components;
- paragraph spacing;
- position of rare forms;
- regularities in empty space.
The filesystem metaphor therefore creates a scanning rule:
Do not equate visual salience with structural importance.
This is another way of breaking the manuscript’s power to steer the observer toward familiar objects.
File Extensions: Surface Labels Can Mislead Type Inference
A filename ending in .jpg looks like an image.
The extension can be wrong.
The data itself must still be inspected.
This is a useful analogy for inherited Voynich section labels.
“Herbal.”
“Astronomical.”
“Biological.”
“Pharmaceutical.”
“Recipes.”
These names are convenient.
They are not demonstrated semantics.
The filesystem tangent gives us a memorable rule:
Do not trust the extension. Inspect the object.
In neutral research language: catalogue labels must not be allowed to become evidence.
File Signatures: Internal Structure Can Identify Type Better Than the Name
Computing systems can identify file types from internal signatures rather than extensions.
The manuscript analogue is powerful.
Instead of classifying a page by what its picture resembles, classify it by a neutral structural signature.
A page signature might include:
- morphology class;
- geometry class;
- hand;
- Currier regime;
- label/prose ratio;
- boundary profile;
- token distribution;
- rare-form profile.
If pages with very different-looking imagery share the same internal signature, visual classification may be overstating difference.
If visually similar pages have different signatures, resemblance may be superficial.
This is the filesystem version of multi-axis classification.
Deleted Files: Absence of an Index Entry Does Not Mean Absence of Data
When a file is deleted, the directory reference can disappear before the underlying storage is overwritten.
This generates a cautious but useful manuscript question.
Could traces of a former organisational structure survive after the explicit navigation system was lost?
Examples might include:
- offset numbering;
- abrupt page transitions;
- broken cross-page patterns;
- partial repetition at quire boundaries;
- contact transfer implying former adjacency;
- damage that truncates a repeated architecture.
The metaphor must not become a fantasy of “deleted chapters.”
The neutral question is narrower:
Can residual physical or structural evidence reveal an organisation whose explicit markers no longer survive?
That is a legitimate reconstruction problem.
Orphan Blocks: Surviving Pieces May Have Lost Their Parent
A damaged filesystem may contain blocks that still hold data but are no longer reachable from the normal directory tree.
This is a different problem from a broken link.
The data survives.
The parent relation is lost.
Voynich may contain analogous structural orphans: a page, label cluster, foldout region or short-entry population whose local structure is coherent but whose broader relationship to the codex is unclear.
Instead of forcing such material into the nearest familiar section, the filesystem tangent recommends an explicit state:
coherent object; parent relation unresolved.
This is a useful research state.
Unknown parentage should not be “repaired” by story.
Filesystem Check: Can Structural Corruption Be Detected Without Knowing Content?
Filesystem check tools inspect metadata and relationships for inconsistencies.
They can detect problems without understanding the meaning of every stored file.
This is one of the strongest transfers to Voynich.
Can we detect likely structural corruption without reading the manuscript?
A Voynich “consistency check” could flag:
- bifolio relationships incompatible with current sequence;
- expected page-pair structures that terminate abruptly;
- hand transitions inconsistent with neighbouring production signals;
- repeated visual structures whose order breaks at damaged boundaries;
- numbering that conflicts with physical collation;
- candidate orphan pages.
None of these flags proves corruption.
They identify where the present object should not automatically be trusted as the original logical sequence.
Before interpreting order, audit order.
Journaling: Later State Can Preserve Evidence of Change
Journaling filesystems record changes so they can recover consistency after interruption.
A medieval manuscript has no digital journal.
But material history can act like a partial change log.
- erasures;
- overwriting;
- repairs;
- different inks;
- paint over text or text over drawing;
- later foliation;
- binding scars;
- ownership annotations.
These traces can sometimes reveal sequence of intervention even when they do not reveal exact dates.
The filesystem tangent therefore asks:
Which surviving marks encode state change rather than original content?
This strengthens the separation between production layers already central to the Voynich Research Library.
Versioning Without Version Control
A filesystem snapshot preserves the state of data at one moment.
The Voynich Manuscript offers no clean snapshots of its own past.
But overlapping physical states can sometimes establish relative chronology.
If drawing lies beneath text, one order is possible.
If pigment covers a line, another relation appears.
If a foliation number crosses later damage, the chronology changes again.
The Tangential question is whether we can reconstruct a partial state sequence:
S₀ → S₁ → S₂ → …
without pretending we know absolute dates for every transition.
This is not a digital snapshot model in the final result.
It is relative stratigraphy of manuscript state.
Defragmentation: A Dangerous Reconstruction Fantasy
Once we imagine fragmentation, we become tempted to “defragment” the manuscript—to rearrange pages until a clean story appears.
This is dangerous.
A computer can defragment because it already knows which blocks belong to which file.
We do not know the Voynich logical structure.
Therefore any proposed reconstruction must be probabilistic and evidence-bounded.
- No rearrangement should be justified by narrative smoothness alone.
- Physical evidence must be prioritised over semantic convenience.
- Several candidate orders should remain alive when evidence does not discriminate.
- Held-out relationships should improve under the proposed order.
- The reconstruction should reduce contradictions globally, not merely fix one favourite section.
Do not reorder the manuscript until the evidence can tell you what “better ordered” means.
Directory Traversal: How Much Does the Chosen Route Change What We See?
Walking a directory tree in one order can foreground different relationships from traversing it another way.
Voynich research does the same thing intellectually.
Start with plants and you collect plant comparators.
Start with astronomy and you collect diagrams.
Start with cipher theory and you collect transformations.
The traversal path changes the evidence neighbourhood we encounter first.
Tangential Voynich therefore asks for traversal robustness.
- Does the same structural population emerge when we begin from hand instead of image?
- From bifolio instead of Currier?
- From boundary behaviour instead of semantic section?
- From labels instead of prose?
If a conclusion depends strongly on the first branch of the analytical tree, we may be measuring search-path bias.
This connects directly to Tangential Voynich’s larger purpose: move the observer so that observer-induced structure becomes visible.
Recursive Descent: A Hierarchy Can Contain Repeating Local Rules
Directories can contain directories, producing nested organisation.
This creates a second hierarchy question after the compiler’s parse-tree experiment.
But now the scale is larger.
A page may contain zones.
A zone may contain entries.
An entry may contain apparent tokens.
A token may contain recurrent components.
The same kinds of boundaries may recur at several scales.
This suggests a multiscale test:
Do similar organisational rules recur at token, line, entry, page and quire scales?
If yes, the manuscript may exhibit scale-repeating organisation.
If no, each scale should be modelled separately.
The filesystem lens forces the scales to confront one another.
Locality: Nearby on the Page, Nearby in the System, or Neither?
Computing distinguishes several kinds of locality.
Data can be physically nearby, logically nearby or frequently accessed together.
These are not equivalent.
Voynich analysis should make the same distinction.
| Locality type | Voynich equivalent |
|---|---|
| Physical locality | Elements near one another on the same page, bifolium or quire. |
| Sequential locality | Elements near one another in present reading order. |
| Structural locality | Elements sharing hand, morphology, boundary or distributional behaviour. |
| Relational locality | Elements repeatedly associated despite physical distance. |
Many arguments silently substitute one locality for another.
Filesystem thinking makes that substitution visible.
Checksums: Can We Build Structural Fingerprints?
A checksum is a compact value used to detect whether data has changed.
Voynich does not contain checksums.
But the concept suggests creating structural fingerprints for pages or page units.
A fingerprint might encode:
- hand distribution;
- Currier classification;
- morphology;
- geometry;
- label/prose ratio;
- boundary profile;
- token-family composition;
- rare-form profile.
Pages with highly similar fingerprints become candidate structural relatives.
Pages that look visually similar but have very different fingerprints become caution cases.
The metaphor disappears.
Multivariate structural fingerprints remain.
Permissions and Ownership: Do Not Confuse Custody With Creation
Files can change owner without changing content.
This is an obvious but useful analogy for manuscript provenance.
A later owner proves custody at a later time.
It does not automatically prove production.
A later foliation can impose a reading path.
It does not automatically reveal original order.
A later note can reveal reception.
It does not automatically reveal original function.
Later control over the object is not the same thing as original authorship of the object.
This keeps the historical provenance firewall intact.
Copy-on-Write: Later Modification Can Preserve Earlier Structure
Some storage systems preserve old data while writing changes into new locations.
The Voynich analogue is not literal copy-on-write.
It is layered modification.
A later hand may add something without erasing the earlier layer.
Paint may alter appearance while leaving underlying drawing visible.
Rebinding may alter sequence while preserving original bifolia.
This yields a strong neutral rule:
Later state should be modelled as transformation of earlier state, not automatic replacement of it.
That matters because the manuscript we see is a stack of historical states.
Reverse Tangential Test: Break a Known Filesystem and Hide Its Labels
The filesystem tangent becomes scientific when we reverse it.
Take a known directory tree.
Remove filenames and directory names.
Preserve only anonymised metadata, block relationships, link structure and access patterns.
Then deliberately damage it.
- Remove directory entries.
- Scatter file blocks.
- Break symbolic links.
- Rename paths.
- Mount one tree inside another.
- Delete some metadata.
Can Voynich-style methods recover:
- logical file groups;
- directory hierarchy;
- fragmented units;
- shared underlying objects;
- broken references;
- mount boundaries;
- orphan blocks;
- probable missing structure?
If not, our confidence in reconstructing unknown codex order should be limited.
Before reconstructing a damaged unknown system, prove that the method can reconstruct a damaged known one.
What Would Make the Filesystem World Fail?
| Filesystem projection | Failure condition | Neutral residue |
|---|---|---|
| Logical vs physical order | Current adjacency predicts structure better than all reconstructed alternatives. | Present order may be largely adequate. |
| Fragmentation | Separated page families do not exceed matched-control similarity. | Repeated template or generic resemblance more likely. |
| Hard-link identity | Repeated forms do not preserve cross-context role. | No stable underlying identity class. |
| Symbolic-link reference | Short forms have no stable relational target. | No reference-like role. |
| Mount point | Apparent boundaries lack aligned independent production signals. | Composite-host hypothesis weakens. |
| Orphan block | Supposed orphan populations are not internally more coherent than controls. | No detached logical unit. |
| File signature | Multivariate page fingerprints do not outperform visual labels. | Neutral signature adds little. |
| Filesystem corruption | Consistency flags do not correlate with independent codicological anomalies. | Present structure may not be corrupted in tested way. |
The filesystem should be discarded wherever it fails.
The reconstruction questions it created can still survive.
The Filesystem Experiment Pack
- Adjacency challenge: compare current folio neighbours against structurally matched non-neighbours.
- Multi-directory classification: represent pages simultaneously by quire, hand, Currier, morphology, geometry and interface.
- Path robustness: test whether the same populations emerge from different analytical starting points.
- Cross-context identity test: measure whether repeated forms preserve structural role across separated contexts.
- Reference-like form scan: test short forms for stable cross-interface targets.
- Broken-link analysis: identify expected counterparts missing near damaged or uncertain regions.
- Fragmentation test: measure whether separated page families preserve unusually strong internal signatures.
- Metadata continuity model: combine bifolio, hand, quire, damage, contact-transfer and page-structure evidence.
- Mount-boundary scan: search for aligned multi-axis discontinuities indicating possible composite production units.
- Neutral page signatures: build multivariate structural fingerprints independent of semantic section labels.
- Orphan-unit register: preserve coherent populations whose parent relationship is unresolved.
- Order consistency check: flag present transitions inconsistent with multiple independent evidence layers.
- State-layer reconstruction: infer relative sequences of drawing, writing, paint, foliation, damage and repair where evidence allows.
- Multiscale boundary test: compare organisational rules across token, line, entry, page and quire levels.
- Locality separation: distinguish physical, sequential, structural and relational neighbourhoods.
- Reverse damaged-filesystem calibration: test whether the methodology can reconstruct a deliberately corrupted known filesystem.
- Held-out Voynich gate: freeze pages before reconstruction parameters are tuned.
- Metaphor removal: rewrite every surviving result without filesystem terminology.
Five Wrong Worlds, One More Precise Manuscript
We now have five active false worlds:
- Railway: movement and routing.
- Operating system: state and control.
- Compiler: representation and legal structure.
- Database: relations and identity across separation.
- Filesystem: logical organisation versus physical placement.
The filesystem reinforces several earlier invariant candidates—boundaries, bridge forms, missingness, multi-axis identity—but adds a distinct family of questions about order.
| Neutral measurement | Database | Filesystem |
|---|---|---|
| Distant structural relation | Join | Logical path / fragmented file |
| Shared underlying identity | Candidate key | Hard link |
| Reference-like form | Foreign key | Symbolic link |
| Expected missing counterpart | Orphan relation | Broken link |
| Unresolved detached structure | Orphan record | Orphan block |
| Higher-order organisation | Schema | Directory hierarchy |
| Physical/logical mismatch | Non-adjacent join | Fragmentation |
| Structural identification | Composite key | File signature / fingerprint |
The overlap is useful.
The difference is even more useful.
The database says distant things can be related. The filesystem says distant things may once have been one logical object.
That is a testable distinction.
Metaphor Removal
Now remove the filesystem.
Delete directories.
Delete paths.
Delete links, blocks, mount points and permissions.
What survives?
- Current physical adjacency should not be assumed to equal original logical adjacency.
- Pages may belong to several valid organisational hierarchies simultaneously.
- Robust structural populations should be discoverable through multiple analytical routes.
- Repeated forms may preserve cross-context identity or reference-like behaviour.
- Expected-but-missing relationships may help identify loss or displacement.
- One logical population may be physically dispersed through the surviving codex.
- Codicological metadata can be combined with text and visual structure to test continuity.
- Neutral identifiers must be separated from later names and semantic labels.
- Some apparent continuities may join originally distinct production units.
- Visually inconspicuous features may be structurally important.
- Residual evidence can preserve traces of earlier organisation after explicit navigation is lost.
- Coherent units can survive even when their broader parent relation remains unknown.
- Present order can be audited for internal inconsistency before being interpreted.
- Physical, sequential, structural and relational locality should be kept separate.
- Multi-axis fingerprints can test page identity without semantic naming.
Every statement survives without filesystem language.
The filesystem has done its job.
World Return
The Voynich Manuscript is not a filesystem.
But the filesystem has taught us to distrust one of the most innocent-looking properties of a book:
where something happens to be now.
A page can be physically adjacent without being logically related.
A page can be physically distant yet belong to the same structural population.
A label can survive after its counterpart disappears.
A coherent unit can survive after its parent relation is lost.
A later numbering system can create a traversal path without proving original order.
A later binding can produce apparent continuity where production history was discontinuous.
The manuscript therefore has at least two geometries.
The geometry of parchment.
And the geometry of relationship.
One of the central Voynich problems may be learning not to confuse the geometry of the surviving object with the geometry of the system that once used it.
That is not a filesystem theory.
It is the residue the filesystem leaves behind.
Tangential Voynich: The Wrong-System Protocol → · Database Experiment → · Compiler Experiment → · Voynich Research Library →
Next Tangent: treat the manuscript as a telecommunications network. Packets, addressing, routing, congestion, error correction, protocol layers, handshakes and dropped signals will force us to distinguish content from transmission behaviour.
NEXT FALSE WORLD · TELECOMMUNICATIONS
The filesystem separated logical organisation from physical placement. Telecommunications now asks whether some of the manuscript’s regularity belongs to framing, context selection, error tolerance and interface constraints rather than ordinary content.
Continue to Tangential Voynich | Treat the Manuscript as a Telecommunications Network →