New independent field guide, published 17 October 2026. Examples are illustrative. The historical directory records linked below are separate source context, not current recommendations or earlier versions of this essay.
A directory reduces a large search space to a smaller collection of possible destinations. That is useful work, but the usefulness depends on what the collection actually promises. An entry might mean that someone submitted a URL, that an editor classified a page, or that a particular property was checked on a particular date. Those are different achievements. Treating all of them as verified quality makes the interface more confident while making the information less precise. This guide develops a practical editorial model for a small resource collection. Its examples describe an invented directory, not current businesses, verified listings or services operated by this site.
Write the inclusion claim in one sentence
Begin by completing a sentence: an item appears in this collection because someone has done what? A defensible answer names an action, not a flattering adjective. For an invented local-history collection, the action might be reviewing whether a page contains digitised material relevant to a stated region. That establishes relevance under a chosen policy. It does not establish that every transcription is accurate, that the publisher owns every image or that the service will remain available. The sentence should be narrow enough that an editor can consistently decide whether the action was actually completed.
This inclusion claim belongs near the collection's introduction and in its maintenance instructions. Otherwise different editors can quietly apply different standards while the interface implies a single uniform process. One might include any plausible submission; another might inspect every linked page; a third might check only whether the server responds. A shared title such as selected resources hides those differences. A short process statement exposes them and gives readers a reason to interpret inclusion correctly. It also makes later improvements measurable: the team can change the action and document which entries have received the stronger review.
Separate the fields by who supplied them
A resource title can come from the destination page, a submitter or the directory editor. A description can repeat a publisher's self-description or summarise independently observed content. Those origins should not be mixed silently. Use separate internal fields for submitted wording, captured page wording and the editor's published description. When a claim is merely reported, attribute it. If an invented museum page says it holds a complete regional collection, a directory can record that the publisher describes it that way without turning completeness into the directory's independently verified conclusion.
An effective record can be modest: identifier, destination URL, display label, subject category, scope note, source of the description and review event. The identifier should survive ordinary wording changes so that maintenance history stays attached to the same record. Do not use the current title as the only identity key; titles can change or be shared by unrelated sites. The extra provenance fields do not need to clutter every public card. They need to exist so that a public statement can be traced to the evidence and action that produced it when a question arises.
Break the word verified into testable properties
A single badge often bundles several claims that require different checks. Availability is not identity; identity is not accuracy; accuracy is not suitability for a particular user. A successful HTTP response can support a narrow availability observation. Matching a publisher's official contact information might support an identity assessment, subject to the method used. Neither establishes every statement on the page. Replace the universal badge with specific, dated properties where they genuinely help. If the collection cannot explain how a property was checked, it should not display that property as an established result.
In a fictional audit, an editor opens a resource and confirms that its named map viewer loads. The useful note is that the map viewer was observed working on the review date under the stated conditions. It would be a leap to label the entire organisation verified or all its maps accurate. This distinction also helps handle disagreement. A later visitor reporting an unavailable download does not necessarily refute the earlier observation of the viewer. The two reports can concern different features, dates or access conditions. Specific claims make those differences investigable rather than adversarial.
Keep editorial judgement distinct from ranking signals
Popularity, traffic and link-related metrics can describe particular measurements, but they are not universal quality scores. A specialised archive may be excellent for a narrow question while attracting few visitors. A widely linked page can be irrelevant to a particular task. If a directory sorts by a metric, identify the metric, its date and the limits of the interpretation. An unexplained best-first order can otherwise convert a convenient sorting rule into an implied endorsement. Alphabetical order is not inherently superior, but its meaning is usually easier for a reader to understand.
Commercial placement adds another distinction. Payment for visibility is a different basis for inclusion from editorial relevance. A hypothetical collection with sponsored entries should make that relationship legible and keep it separate from any claimed review process. Do not invent a paid service merely because an old site had submission controls. This edition's historical catalogue records old names, addresses and categories without reviving rankings or submissions. Its captured PageRank references are not current measurements and are not republished as badges. Preserving an old record does not require preserving the persuasive force of its original interface.
Design review around changes that matter
Maintenance begins with a question: what could change that would make this entry misleading? The destination could move, the subject could change, access could become restricted or a once-specific description could become obsolete. Different changes require different responses. An address redirect may call for a URL update after content review; a change of subject may require removal from the current collection. A temporary server failure should not automatically erase historical evidence. Record the observation first, then apply a defined editorial decision. That separation prevents a transient technical condition from silently rewriting the collection's meaning.
Review effort should match the collection's purpose. A directory of stable public documents may need a different schedule from one pointing to frequently changing services. Use risk and change history to prioritise checks, and distinguish a planned next review from a completed one. A calendar entry saying review next month does not mean the destination has already been checked this month. In an invented workflow, an editor can queue uncertain entries, preserve the last supported description and publish a clear limitation while waiting for evidence. The limitation is useful information, not an excuse to replace inspection with a confident guess.
Publish a collection readers can calibrate
The public interface should answer three questions quickly: what is this resource about, why is it included and what has actually been checked? A concise description and a visible scope statement can do more than a wall of badges. Link text should help readers anticipate the destination; W3C's link-purpose guidance gives a useful accessibility foundation for that goal. The original editorial decisions proposed here go further by tying each published claim to a defined action. Accessibility, relevance and verification are related concerns, but none automatically supplies the others.
A directory succeeds when it helps someone choose a promising next step with the right expectations. It does not need to pretend to certify the entire web. The historical records linked below preserve an earlier directory's identity and submission-policy context; they are not active rules for this edition. The practical model is straightforward: define inclusion, preserve provenance, name specific checks and keep maintenance events dated. Readers can then use the collection as a map of leads while still making the additional judgements their own task requires. Precision about that boundary is part of the directory's value.
References & further reading
From the recovered archive
- Submission rules — historical policy · captured 28 November 2010
- Computers & Internet / Web Directories · captured 25 December 2010