Present a digital object without trapping it inside one viewer
The International Image Interoperability Framework separates the intellectual object, its presentation surfaces, content resources and annotations so repositories and viewers can exchange meaning as well as pixels.
A scanned book, painting, map or manuscript is rarely just one image. It has ordered pages or surfaces, labels in several languages, rights information, thumbnails, transcriptions, commentary and links to services. A proprietary viewer can display all of that while making the structure difficult to reuse.
IIIF Presentation API 3.0 provides a shared JSON-LD model. A Manifest describes one object; Canvases provide ordered presentation spaces; Annotation Pages group annotations; and painting annotations place content on a Canvas. The model works best when each layer has a clear identity and verified role.
Chapter contents
- 1. Choose the IIIF resource that matches the object
- 2. Give the Manifest a stable identity and honest boundary
- 3. Treat a Canvas as an abstract presentation surface
- 4. Use Annotation Pages to organise statements about a Canvas
- 5. Understand target, body and motivation
- 6. Worked example: publish a two-page letter
- 7. Connect the Image API without confusing the layers
- 8. Model order, structure and navigation with Ranges
- 9. Write multilingual labels and metadata for people
- 10. State rights, provider and required statements accurately
- 11. Validate JSON-LD, API requirements and viewer behaviour separately
- 12. Version manifests without breaking citations and annotations
- 13. Audit interoperability from discovery to reuse
Choose a reading route
- Route 1: Orient the object — Collection, Manifest and boundary
- Route 2: Build surfaces — Canvases, dimensions and order
- Route 3: Work a manifest — painting annotations and services
- Route 4: Add meaning — metadata, rights and multilingual labels
- Route 5: Publish and audit — validation, change and transfer
How to use this guide
Start with the chapters that match your immediate decision, then return to the worked example and audit chapters. Each chapter separates a model from the evidence needed to use it responsibly. A claim is not strengthened merely by being technical; it is strengthened when assumptions, observations, alternatives and limits can be inspected.
| Layer | Question | Minimum record |
|---|---|---|
| Concept | What exactly is being represented? | Definitions and boundary |
| Method | How will the reasoning be carried out? | Steps, inputs and assumptions |
| Transfer | Would it work in a changed setting? | A deliberately altered case |
| Audit | What could prove it wrong? | Evidence, limits and revision trigger |
Chapter 1. Choose the IIIF resource that matches the object
The idea
A Collection groups resources; a Manifest describes one compound digital object; a Canvas models one presentation surface. Correct hierarchy keeps discovery separate from page-level delivery.
A disciplined method
Name the intellectual object and intended user task, decide whether the top resource is a Collection or Manifest, then inventory ordered surfaces.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is whether several items form one object or a discovery grouping. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A library Collection lists three manuscript Manifests; each Manifest has ordered Canvases for its pages.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Change the case to a multi-volume work and decide whether volumes are Manifests within a Collection or one Manifest with ranges.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: The model cannot settle curatorial identity without a documented local policy.
Evidence to retain: object inventory, hierarchy rationale, identifiers and user tasks
Common misconception: Every folder should become a Collection and every image a Manifest.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 2. Give the Manifest a stable identity and honest boundary
The idea
A Manifest represents the presentation of one object, not the viewer page or image server. Its identifier should remain stable while implementations and services change.
A disciplined method
Mint an HTTPS identifier, provide type, label and ordered items, record provider and rights where known, and state what is excluded.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is which digital resources belong to this presentation. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A map Manifest includes front, reverse and detail Canvases but links the catalogue record as seeAlso context.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Move the viewer to a new domain while keeping the Manifest identifier and object boundary stable.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: Stable URLs require governance; a Manifest can be incomplete or one of several valid presentations.
Evidence to retain: identifier ownership, redirect policy, boundary note and release record
Common misconception: The Manifest URL should be the current viewer URL.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 3. Treat a Canvas as an abstract presentation surface
The idea
A Canvas is a coordinate space onto which content is painted. It can represent a page, painting surface, audio duration or other view without being identical to one image file.
A disciplined method
Create one Canvas per meaningful surface, assign dimensions or duration appropriate to the media, and keep the ordered sequence explicit.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is what physical or conceptual surface the Canvas stands for. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A folio Canvas is 3000 by 4200 units and can carry a full image plus a transcription region.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Replace the scan with a higher-resolution image while retaining the Canvas identity and coordinates.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: Canvas dimensions are a presentation coordinate system, not necessarily physical measurements.
Evidence to retain: surface inventory, order, dimensions, source-image metadata and mapping tests
Common misconception: A Canvas is merely an image URL.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 4. Use Annotation Pages to organise statements about a Canvas
The idea
In Presentation 3.0, Canvases include Annotation Pages, which contain annotations. The page provides ordering and a retrieval boundary for related annotations.
A disciplined method
Create the Canvas items array, add Annotation Page identifiers and types, and place annotations on the page rather than directly on the Canvas.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is which annotations belong together and must load together. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
One Annotation Page carries the main painting annotation; another external page carries transcription annotations.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Change the annotation source from embedded to external and verify that order and targets remain usable.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: Very large annotation pages need pagination or segmentation strategies beyond a small example.
Evidence to retain: page identifiers, ordering, item counts, retrieval tests and provenance
Common misconception: Annotations can be inserted directly into a Canvas items list.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 5. Understand target, body and motivation
The idea
A Web Annotation connects a body to a target with a motivation. For core display, a painting annotation targets a Canvas and has an image, audio or video body.
A disciplined method
Check each annotation type, motivation, body format and target; use selectors or fragments only when the body applies to part of a Canvas.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is whether content covers a whole surface or a defined region. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A JPEG body paints the full Canvas; a textual annotation targets a rectangular region and uses a commenting or supplementing motivation.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Transfer the annotation to an audio Canvas and replace spatial coordinates with a temporal fragment.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: Viewer support for complex selectors and motivations varies; interoperability must be tested.
Evidence to retain: expanded annotation graph, body media type, target resolution and viewer behaviour
Common misconception: Any link between content and a Canvas is a painting annotation.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 6. Worked example: publish a two-page letter
The idea
A small Manifest can be inspected completely and used to verify the relationships among Manifest, Canvas, Annotation Page, annotation and image service.
A disciplined method
Create one Manifest, two ordered Canvases, one painting Annotation Page per Canvas and one Image API service per image body. Add labels, provider and rights.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is how content delivery remains separate from presentation structure. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A fictional letter has recto and verso Canvases. Each points to a painting annotation whose body references a JPEG and a level-2 image service.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Change the second page to a foldout with a different aspect ratio and confirm viewers preserve its Canvas dimensions.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: The fictional letter demonstrates structure, not cataloguing completeness or rights clearance.
Evidence to retain: saved JSON-LD, image-service responses, validator output, screenshots and identifier log
Common misconception: A working image viewer proves that the Manifest is conformant.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
| Resource | Role | Example |
|---|---|---|
| Manifest | One letter | letter/manifest |
| Canvas | One surface | page-1/canvas |
| Annotation Page | Ordered annotation set | page-1/painting |
| Annotation | Places content | page-1/image-annotation |
| Image service | Delivers derivatives | iiif/2/image-id |
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 7. Connect the Image API without confusing the layers
The idea
Presentation describes what is shown and in what order; Image API services deliver requested regions, sizes, rotations, qualities and formats. The image service supplements an image body.
A disciplined method
Describe the image body with dimensions and format, attach the service with its profile, and test info.json plus representative image requests.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is which server owns image delivery and which resource owns presentation. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
The Manifest references a JPEG body with an Image API 3 service; the viewer chooses a viewport-sized derivative.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Switch the image server while retaining the Manifest, Canvas and annotation identities; document the service change.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: Image service capabilities differ by compliance level, and a service outage still breaks display.
Evidence to retain: service identifier, profile, info.json, request tests and cache behaviour
Common misconception: A Manifest replaces the Image API or should encode every derivative URL.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 8. Model order, structure and navigation with Ranges
The idea
Canvas order gives the basic sequence. Ranges provide logical structures such as chapters, articles, scenes or table-of-contents entries without changing Canvas identity.
A disciplined method
Keep items in reading order, create labelled Ranges for meaningful sections, nest cautiously and test links to Canvases or temporal/spatial parts.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is whether the structure is physical order or interpretive navigation. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A book Manifest orders 120 Canvases and provides Ranges for preface, chapters and index.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Change one chapter to span only part of an audio Canvas and verify the temporal target.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: Multiple legitimate structures may exist; one Range tree should not erase alternative readings.
Evidence to retain: sequence tests, Range graph, labels, targets and curator review
Common misconception: Ranges are required to make a sequence of pages valid.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 9. Write multilingual labels and metadata for people
The idea
IIIF language maps let labels, summaries and metadata values be provided in one or more languages. Values should aid discovery without pretending free text is controlled data.
A disciplined method
Use valid language tags, provide a none entry only when language is genuinely unknown or inapplicable, and keep metadata labels concise and ordered.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is which languages readers need and which value is authoritative. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A Manifest supplies English and Malay labels while retaining an untranslated shelfmark under none.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Add a right-to-left label and inspect direction, font and fallback in more than one viewer.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: Language maps do not translate automatically and metadata entries are presentation hints, not a full catalogue schema.
Evidence to retain: language tags, source translations, reviewer names, fallback tests and catalogue links
Common misconception: The none language is a convenient substitute for every untranslated value.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 10. State rights, provider and required statements accurately
The idea
Rights information tells users the status and conditions that apply to the represented object or supplied content. Provider and requiredStatement support attribution and institutional context.
A disciplined method
Use a recognised rights URI where applicable, write a visible required statement, distinguish copyright from access restrictions and confirm body-level exceptions.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is which entity and layer the statement governs. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A public-domain artwork Manifest identifies its provider and required credit while a modern photograph body carries separate rights.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Change the image source to a licensed partner and identify which statements must change.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: A rights URI is not legal advice and cannot resolve unknown ownership or privacy restrictions.
Evidence to retain: rights assessment, licences, credit text, takedown route and body exceptions
Common misconception: One Manifest-level rights value automatically governs every linked resource.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 11. Validate JSON-LD, API requirements and viewer behaviour separately
The idea
A document can parse as JSON yet violate the Presentation API, and a conformant Manifest can expose a viewer bug or inaccessible content.
A disciplined method
Validate syntax and schema constraints, expand JSON-LD, check every identifier and referenced service, then exercise several viewers and assistive technologies.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is which failure belongs to data, service or client. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A Manifest validates but one Annotation Page returns HTML; a viewer silently skips it while a link checker catches the media-type error.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Move external annotation pages behind slower hosting and test loading, timeout and graceful fallback.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: No validator can confirm curatorial truth, accessibility quality or long-term persistence.
Evidence to retain: validator version, HTTP traces, viewer matrix, accessibility review and exception log
Common misconception: If one popular viewer renders the object, the Manifest is interoperable.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 12. Version manifests without breaking citations and annotations
The idea
Changing labels is different from changing Canvas order or identity. Annotation targets and bookmarks depend on stable identifiers and coordinates.
A disciplined method
Define change classes, preserve identifiers where meaning is stable, issue new versions for boundary or sequence changes, and redirect deprecated resources deliberately.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is which changes preserve the same Canvas and annotation target. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A corrected summary updates in place; splitting one Canvas into two triggers a migration map for external annotations.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Transfer a Manifest between repositories and maintain old identifiers through redirects or sameAs only when equivalence is justified.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: Versioning practices vary, and careless sameAs claims can merge non-equivalent presentations.
Evidence to retain: change log, previous JSON, redirect tests, annotation migration and citation policy
Common misconception: Because JSON is dynamic, every edit may safely overwrite the previous Manifest.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Chapter 13. Audit interoperability from discovery to reuse
The idea
Interoperability is successful when another system can discover the Manifest, retrieve its resources, understand the structure and reuse content under stated rights.
A disciplined method
Test direct retrieval, content types, CORS, identifiers, language maps, image requests, annotations, ranges, rights and offline preservation; record the receiving client.
A useful working record makes the sequence visible: first state the question and unit of analysis; next record the observations and their provenance; then apply the stated rule; finally compare the result with plausible alternatives. For this chapter, the decisive choice is which minimum route must work without the original viewer. Keeping that choice explicit prevents a later reader from mistaking a convenient convention for a fact about the world.
Worked example
A second institution imports the Manifest, displays both pages, follows Image API services and preserves attribution without custom code.
Work the example twice. On the first pass, follow the stated assumptions. On the second, reverse one assumption, remove one piece of evidence and ask whether the conclusion, ranking or boundary changes. This small stress test matters because polished outputs can hide sensitivity. The objective is not to force certainty; it is to locate the evidence on which the conclusion actually turns.
Changed-context transfer check
Change the receiving client to one that implements only a subset of Presentation 3.0 and document graceful degradation.
A transfer succeeds only when the structure of the problem survives the change. Match roles, dependencies and measurement conditions rather than surface vocabulary. Record what was preserved, what changed, and why the method still applies—or why it does not. If the new context changes the data-generating process or the people bearing the burden, re-specify the model instead of importing the old answer.
Limits, evidence and revision
Limit: Shared standards reduce custom integration but do not guarantee identical interfaces, performance or preservation.
Evidence to retain: end-to-end test, client versions, network trace, screenshots, rights display and reuse outcome
Common misconception: Publishing JSON at a URL is the same as achieving interoperability.
A gold-standard conclusion is therefore conditional. It should name the observation period, excluded cases, uncertainty and the event that would trigger review. Negative findings and unresolved contradictions belong in the record. They protect future readers from repeating a hidden assumption and make correction possible without reconstructing the entire analysis from memory.
Practical evidence check
Can another reader recover the input, reproduce the intermediate decision and identify at least one observation that would change the result? If not, the chapter has produced an explanation but not yet an auditable finding.
Field workbook
Create a two-Canvas Manifest for a fictional or rights-cleared object. Draw the resource graph first, then validate the JSON, open it in two viewers and record which labels, annotations, services and rights statements survive.
- Write the decision question in one sentence and name the decision owner.
- List observations separately from interpretations, with dates and provenance.
- Record at least two plausible alternatives before choosing a preferred account.
- Run the changed-context test and note which assumptions fail.
- State the limit, confidence and next review trigger in plain language.
| Layer | Evidence | Revision trigger |
|---|---|---|
| Object boundary | Manifest inventory | Pages added or split |
| Presentation | Canvas order and Ranges | Sequence changes |
| Content | Annotation bodies and services | Server or media change |
| Reuse | Independent client test | Client cannot interpret |
Continue through the verified ecosystem
- OAI-ORE resource maps — compound object boundaries and contextual identities
- RO-Crate research objects — portable data and contextual metadata
- Reference data and crosswalks — version labels and mappings
- Data citation — versioned citation and fixity
- PREMIS preservation evidence — preservation events and custody
Frequently asked questions
What does a IIIF Manifest represent?
One compound digital object and its ordered presentation, not the viewer application itself.
Is a Canvas an image?
No. It is an abstract presentation surface onto which images, audio, video or text can be placed with annotations.
Why are Annotation Pages needed?
Presentation 3.0 groups ordered annotations in Annotation Pages, which Canvases reference through their items.
What does painting mean?
It is the motivation used when an annotation places content such as an image on a Canvas for display.
Does IIIF settle copyright?
No. It carries rights and attribution information, but institutions still need a sound rights assessment.
How do I test interoperability?
Validate the Manifest, test every referenced service, and open it in more than one independent client.
Authoritative sources and further reading
- IIIF Presentation API 3.0 — authoritative presentation specification
- IIIF Image API 3.0 — region, size, rotation, quality and format requests
- Presentation API 3.0 change log — differences from earlier versions
- W3C Web Annotation Data Model — body, target and motivation model
- IIIF Cookbook — community implementation recipes
This guide explains IIIF information architecture. It does not clear rights, certify a repository or replace the exact API specification and implementation guidance used by your institution.