Resources · 75

Network failures: preserve input and resume a task

Describe waiting, failure and uncertain outcomes without losing the person’s work.

· 3 min

Communications racks and network cabling in an empty equipment room AI-generated editorial illustration

What this guide helps achieve

  • Describe states
  • Preserve proportionately
  • Explain recovery
  • Replay interruptions
  • Check the request before repeating it

Quick check

  • Does useful input survive interruption?
  • Can the person distinguish waiting from confirmation?
  • Can retrying create a duplicate effect?

Step-by-step method

  1. 01

    Describe states

    Separate local input, started submission, confirmed receipt and uncertain outcomes. Missing responses can occur after server receipt. Base confirmation on business state rather than a visual change alone.

    Deliverable: state map.

  2. 02

    Preserve proportionately

    Keep useful fields during errors and indicate what reload will lose. Decide explicitly whether drafts may be stored, where and for how long. Avoid automatically retaining sensitive data on shared devices.

    Deliverable: draft policy.

  3. 03

    Explain recovery

    Provide understandable messages available to assistive technologies and a specific action. Distinguish fixing a field, checking a request and retrying. Keep focus and navigation coherent; color alone or fleeting toasts are insufficient.

    Deliverable: messages and keyboard journey.

  4. 04

    Replay interruptions

    Test offline before submission, interrupted transfer, delayed response and restored connectivity. Use fictional data and verify retention, announcements, focus and absence of duplicated effects. Include reload and browser-back behavior.

    Deliverable: end-to-end recovery evidence.

  5. 05

    Check the request before repeating it

    Tie recovery to the request and the person authorized to inspect it. Define how long the identifier remains usable, behavior after expiry and a support path when confirmation remains unavailable. A retry button does not establish server-side uniqueness.

    Deliverable: status-check contract and expiry scenario.

Choose an action from what is actually known

Fictional scenarios for recovery planning. The interface response and server receipt are distinct observations.

ObservationAction to prepareExpected evidence
Connectivity disappears before submissionPreserve input allowed by policy and explain how to resume.No operation created in this scenario; allowed fields still available.
The server accepts but its response is lostExplain the uncertain outcome and inspect the existing request before another attempt.Same business identifier, recovered result and no additional effect.
A field is rejected with a confirmed errorIdentify the field and cause in text, preserve other values and allow correction.Message associated with the field, keyboard journey and validated correction.
The session expires during recoveryRestore an authorized session before displaying status; apply draft policy on shared devices.No cross-account access to requests or drafts and a follow-up path after expiry.

Reusable worksheet

Complete with your authorised observations. These fields are a working template, not observed results.

FieldInformation to record
StateLocal, sent, confirmed or uncertain
RetentionFields, location, duration and deletion
RecoveryMessage, focus, verification and business effect
AcceptanceVerified business state, one effect, available input and announced message

Worked example

Illustrative situation

Fictional example: a request form remains waiting after a lost response.

Decision and expected evidence

The interface keeps fields, announces uncertainty and checks the request identifier before retrying. One case is created and confirmation is accessible.

Distinguish the mechanisms

MechanismPurposeCheck or limitation
Confirmed errorOffer correction or another attemptCheck whether an operation already exists
Uncertain outcomeLook up request stateDo not announce failure as an established fact

Management indicators

IndicatorWhat it measuresFirst action
Recovered inputUseful fields retained in defined casesFix losses and excessive retention
Resumed tasksOne confirmed outcome after simulated failureReview messages, state checks and repeats

Common pitfalls

  • Erase input after a transient error
  • Retry an external effect without checking its state

Frequently asked questions

Does navigator.onLine establish service access?

No. Connectivity indications do not prove your server responds or an operation succeeds.

Should all forms be stored locally?

No. Choose fields and durations according to sensitivity, device and actual recovery needs.

Can retries be automatic?

It depends on the operation. External effects require repeat protection and appropriate state checks.

Should a status announcement always move focus?

No. A status added without a change of context can be announced to assistive technologies without receiving focus. An error requiring correction must remain identifiable; also test navigation and avoid repetitive announcements.

Official references

References consulted: . The method and worksheet propose checks to adapt to your context; they do not constitute certification.