Research Guides
Keep the Base URL With an Exported Link
A link exported from page research may be only a path, not a complete destination. If the receiving worksheet forgets where that path came from, a later reviewer can check the wrong address and report a broken citation that never existed on the original page.
This workflow preserves the context needed to interpret relative references. It is a manual research method, not a claim about an automated Colquhoun.shop URL resolver.
1. Keep the original reference and its context
For every extracted link, retain the source page address, the reference exactly as observed, the observation time and the resolved destination in separate fields. Do not overwrite the original reference with a cleaned-up guess.
Record the effective base used for resolution. When working from a browser, inspect the actual document context rather than assuming the URL originally typed is always sufficient. If that context is unavailable in an imported file, flag the row for investigation.
2. Resolve rather than concatenate
MDN explains that resolving a relative reference against a base URL is not simple string concatenation. The URL API accounts for references relative to the current directory, a parent directory or the site root. The trailing slash in a base can therefore affect the result. MDN: Resolving relative references to a URL
Use a standards-aware resolver in a local test, with the base explicitly supplied. Preserve errors as errors; do not turn an invalid reference into an apparently valid address by guessing which characters to remove.
3. Test a tiny fixture before processing the export
Consider the invented reference methods. With the base https://example.com/reports/, it resolves to https://example.com/reports/methods. With https://example.com/reports, it resolves to https://example.com/methods. These are different addresses, even though a casual glance might treat the base strings as equivalent.
Keep those cases in a small regression fixture alongside the kinds of references actually present in your dataset. Run the fixture after changing import or normalization logic. No network request is needed merely to calculate the result.
Do not infer which address is the intended destination from this calculation. Resolution establishes the address produced by the supplied context, not the publisher's editorial intent or the existence of a page there.
4. Separate resolution from destination verification
Once the destination is resolved, review whether it is appropriate to visit and then inspect it within your authorized scope. Record any redirect separately from the resolved starting address. A successful response does not establish that the destination supports the claim beside the link.
Google's link guidance recommends anchor elements with an href that resolves to a real web address. It warns that script-event navigation without suitable link markup is not a reliable source of extractable URLs for its crawlers. Google: Link best practices
If an export captured only an event handler or an incomplete fragment of markup, mark the extraction limitation. Do not manufacture a destination to make the report appear complete.
5. Hand off a reproducible link record
- Source: page and observation time.
- Original: exact reference captured.
- Context: effective base and how it was established.
- Resolution: calculated address or explicit error.
- Verification: visited address, redirect and content assessment.
- Limits: missing context or unvisited destinations.
Keep resolution and later verification timestamps separate if they happened at different times. That makes it possible to distinguish a processing mistake from a destination that changed after the research.
The final deliverable should let another reviewer reproduce the address calculation before repeating the content check. That is stronger evidence than a single column of URLs whose relationship to the original pages has been lost.