Lecture hypotheses and ways to test them

Web development: past, present, and possible futures · 2026

This student-facing summary states the proposals used in the lecture. It is not an excerpt from, or empirical validation of, the unpublished position paper cited in the background material.

Explicit actions may reduce client guesswork

Give an action a name, valid inputs, preconditions, an outcome and failure behavior. Test whether people, assistive tools and agent clients can complete the same task with fewer errors. Their needs differ; shared semantics alone establish neither accessibility nor agent reliability.

A generated view may help a specific task

Compare a fixed view and a constrained generated view using the same source revision and task. Record completion, mistakes and unsupported claims. Extra variability, invented facts, inaccessible output or unauthorized controls count against the proposal. Generating a view does not demonstrate safe arbitrary service execution.

A context receipt may make output easier to inspect

Record selected facts, aggregate revision, data destination and resulting view or action. Check whether a reviewer can trace one displayed claim to its evidence and identify missing support. A receipt is an audit aid, not proof of correctness or a security standard.

Boundary used in the demo

Use reviewed seminar facts and predefined-field aggregates, excluding private questions and identifiers. Validate output structure, allowed components and URLs. Keep a deterministic fixed view when generation is unavailable or invalid. Permit only the existing explicitly authorized actions.

Further context

Hypermedia as a Substrate for the Agentic Web: From Documents to Affordances (2026) is an unpublished position paper supplied by the lecturer. Students can use this summary without access to that manuscript. These proposals remain hypotheses until evaluated against the evidence above.