Data Island Integration: Embed RDF in HTML with solidos-lite #7

Summary

Integrate solidos-lite to embed RDF data directly in HTML responses, eliminating the need for a second HTTP request when viewing resources in the browser.

Current Behavior

Browser → GET /resource (Accept: text/html)
Server  → Returns HTML wrapper that loads mashlib from CDN
Browser → mashlib initializes, fetches /resource AGAIN (Accept: text/turtle)
Server  → Returns RDF data
Browser → mashlib renders the data

Problems:

• Two HTTP round-trips required
• Cannot work offline after initial load
• Not suitable for static HTML exports
• Latency penalty for every resource view

Proposed Behavior

Browser → GET /resource (Accept: text/html)
Server  → Returns HTML with:
          - Embedded RDF in <script type="text/turtle"> data island
          - solidos-lite shim that intercepts mashlib fetcher
Browser → solidos-lite patches fetcher.load()
Browser → mashlib reads from data island (no network request)
Browser → Renders immediately

How solidos-lite Works

solidos-lite intercepts mashlib RDF fetching by patching fetcher.load:

const originalLoad = fetcher.load.bind(fetcher)

fetcher.load = async function(uri, options = {}) {
  const island = findDataIsland(uriStr)
  
  if (island) {
    // Parse locally embedded data instead of fetching
    $rdf.parse(content, store, baseUri, contentType)
    fetcher.requested[baseUri] = "done"
    return doc
  }
  
  // Fall back to original network fetch
  return originalLoad(uri, options)
}

Key interception points:

1. fetcher.load() - main fetch function
2. fetcher.nowOrWhenFetched() - callback-based fetching
3. fetcher.getState() - reports "fetched" for data island URIs

Implementation Plan

Phase 1: Basic Integration

File: src/mashlib/index.js

Update generateDatabrowserHtml() to:

1. Accept RDF content as parameter
2. Embed content in <script type="text/turtle" data-uri="..."> tag
3. Load solidos-lite from CDN before mashlib

Phase 2: Handler Integration

File: src/handlers/resource.js

Pass RDF content to HTML generator when serving mashlib view.

Phase 3: Content Type Handling

• text/turtle → embed directly
• application/ld+json → embed as JSON-LD
• text/n3 → embed directly
• application/rdf+xml → fallback to fetch

Phase 4: Large Resource Handling

const MAX_EMBED_SIZE = 100 * 1024; // 100KB threshold

if (rdfContent.length > MAX_EMBED_SIZE) {
  return generateDatabrowserHtml(resourceUrl); // fetch-based
} else {
  return generateDatabrowserHtml(resourceUrl, rdfContent, contentType); // embedded
}

Benefits

Considerations

Security

• Embedded content must be properly escaped to prevent XSS
• RDF in <script> tags is generally safe (not executed as JS)

Caching

• Browser caching differs for embedded vs fetched content
• Consider ETags/Last-Modified for HTML response

Linked Resources

• solidos-lite falls back to network fetch for linked resources
• Only primary resource is embedded

Testing Plan

1. Unit Tests: generateDatabrowserHtml() with embedded content
2. Integration Tests: Verify single network request in devtools
3. Manual Tests: Save HTML, open locally, verify offline rendering

References

• solidos-lite repository
• solidos-lite npm
• mashlib repository

Related Issues

• Issue Add did:nostr SSO and Schnorr (NIP-98) authentication #4: did:nostr authentication (completed)
• Issue Phase 2: Link did:nostr to WebID profiles (owl:sameAs) #6: WebID linking for did:nostr

LLM-friendly issue for future implementation.