VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Portability Of Civilisation | Coordinating Commitments: The Space Between Two Successful Steps

A joint commitment is easiest to describe when one coordinator controls all the relevant records. Many real arrangements do not have that boundary. A kit reservation may succeed in one system before a slot request fails in another. A message may arrive while its acknowledgement is lost. The space between two successful steps is itself part of the mechanism we need to carry.

This chapter develops a proposed coordination contract for that space. It uses fictional session plans to distinguish pending work, confirmed commitments, uncertain outcomes and compensating actions. Composition Lab’s checked implementation remains a single sequential, in-memory model. The multi-service arrangements below are explanatory designs, not implemented distributed software or reports of a real service.

  • Two individually successful operations can still leave the overall purpose unfinished.
  • A missing reply does not establish that the requested action failed.
  • Pending, rejected, compensated and unknown describe different situations.
  • Compensation is a new action with its own conditions and possible failures.
  • The host must decide when a partial result may be offered and when activity must pause.
  • A visible workflow can help people and AI coordinate without disguising who owns each decision.
  • A monetary receipt, where relevant, settles only the claim it actually represents.

Book and reading map · Previous chapter · Next chapter

eduKate · CivilisationOS · AI and Education · Book Chapter 31 · Part 6 · Packet route chapter 12

Choose your reading route

Read in order, or choose a route through this chapter. All sections remain open below.

Open the section index

Section index

Sections 1–6 · Purpose, pending work and rejection
  1. Leave the convenient boundary deliberately
  2. Define a purpose above the individual actions
  3. Begin with an honest pending state
  4. A missing acknowledgement creates uncertainty
  5. Retrying requires an agreement about identity
  6. Rejection leaves work to do
Sections 7–12 · Compensation, records and ordering
  1. Compensation is a new operation
  2. A compensating request can remain unresolved
  3. Use a small state vocabulary with clear meanings
  4. Keep an obligation record beside the event record
  5. Three traces can end with the same visible message
  6. Order the steps according to their consequences
Sections 13–18 · Time, successors, AI and autonomy
  1. Time limits require an owner and a clock
  2. A handover must carry unfinished plans
  3. Distributed ownership need not mean absent responsibility
  4. AI can coordinate messages without owning the purpose
  5. Monetary settlement is another bounded step
  6. Preserve local autonomy during interruption
Sections 19–24 · Interruption, matching evidence and closure
  1. Test interruption at every boundary
  2. The connection needs its own packet
  3. A matching identifier is necessary but insufficient
  4. Define when further attempts should stop
  5. Manual coordination is a legitimate receiving route
  6. Sources and proposal boundary

SECTION 1 OF 24

1. Leave the convenient boundary deliberately

Back to contents

The preceding chapter placed two resource maps inside one operation. That allowed the teaching model to validate both before changing either. Now suppose the kit mechanism and the session mechanism belong to different owners. Each maintains its own records and decides whether to accept a request. The caller cannot simply rewrite both states together.

This separation may be entirely reasonable. A lending team and a teaching team may have different responsibilities, opening hours and review processes. Requiring them to become one organisation would be a substantive institutional change, not a small software convenience. The architecture should begin with the actual ownership of the decisions.

The resulting coordination problem is not solved by drawing an arrow between the packets. The arrow must carry a defined request, and the receiver must return a result whose meaning the sender understands. When the exchange stops partway through, somebody must know what is still owed, what can be released and what remains uncertain. Those responsibilities belong in the transferable design.

Contents · Next section

SECTION 2 OF 24

2. Define a purpose above the individual actions

Back to contents

The fictional purpose is to prepare a session with one suitable kit and one suitable slot. Reserving a kit is a step towards that purpose. Reserving a slot is another step. If the caller treats either success as completion, it can tell the learner the session is ready before the required combination exists.

A coordination record therefore needs an identity for the overall plan, distinct from the identities of the individual requests. Plan P can refer to a kit request K and a slot request S. The relationship lets a successor discover which local operations were intended to support the same purpose.

The relationship should also preserve scope. A successful kit request does not imply a successful slot request merely because both carry P. The shared plan identity helps assemble evidence; it does not manufacture missing evidence. The overall status must be derived from the relevant local outcomes and the declared completion rule.

Contents · Previous section · Next section

SECTION 3 OF 24

3. Begin with an honest pending state

Back to contents

After the kit is held but before the slot is confirmed, the plan is pending. That state can be useful: it tells the host that work has begun and identifies the unresolved dependency. Calling the plan ready would overstate the evidence. Calling it failed would discard a commitment that may still be valid and useful.

The pending state should say what it is waiting for. A vague spinning indicator does not help a successor distinguish a delayed reply from an unavailable resource. A useful record names the outstanding request, the owner expected to answer and the next permitted step. The user-facing explanation can remain simple while the underlying record stays precise.

For a classroom role-play, a learner can hold a card labelled kit confirmed while the slot card remains unanswered. The group then decides which statements it may truthfully make to the intended participant. This makes pending work visible as a substantive condition, rather than treating it as an embarrassing gap to be covered with optimistic language.

Contents · Previous section · Next section

SECTION 4 OF 24

4. A missing acknowledgement creates uncertainty

Back to contents

Suppose the slot owner accepts the request, but the acknowledgement never reaches the coordinator. The coordinator sees silence. From that observation alone it cannot distinguish a request that never arrived from one that committed successfully. A timeout identifies a limit in waiting; it does not reveal which history occurred.

The proposed next step is to investigate using the request identity and the receiving system’s declared status route. If the receiver can report the earlier committed outcome, the coordinator may resolve the uncertainty. If no such route exists, the plan may remain unresolved and require a person to reconcile the records.

This is a different problem from a clear rejection. A rejected slot request tells the coordinator that this request did not obtain the slot under the receiving system’s contract. An unanswered request does not supply that same information. A portable workflow must preserve the difference because the correct response to each can be different.

Contents · Previous section · Next section

SECTION 5 OF 24

5. Retrying requires an agreement about identity

Back to contents

Repeating an uncertain request can be appropriate if the receiver recognises the repeated identity and preserves the original result. The earlier lab demonstrates this idea within one in-memory state. A real service would need a durable, appropriately scoped implementation and an explicit rule for how long it retains the relevant history.

The sender and receiver must agree on what makes a request the same. Is the identity unique within one plan, one host or the entire service? What happens if the same identity arrives with different contents? Can a successor discover the original payload? These questions determine whether retrying is a way to recover information or a way to create an accidental second commitment.

Our proposed multi-service workflow assumes no guarantee that has not been provided by the receiving component. If a component lacks a reliable repeat-request contract, the coordinator should not invent one. It may need to stop, query a human owner or use a different procedure. The inability to automate a step cleanly is a dependency to record.

Contents · Previous section · Next section

SECTION 6 OF 24

6. Rejection leaves work to do

Back to contents

Now consider a clear rejection: the kit request succeeded, but the slot owner says no. The overall plan cannot be presented as ready. Yet the kit is still held. If the coordinator abandons the plan without further action, it may leave a resource unavailable to others for no useful purpose.

The workflow needs a declared response. It might ask for another suitable slot, ask the requester whether an alternative is acceptable, or release the kit. These choices depend on the original mandate. A coordinator authorised to reserve a particular combination is not automatically authorised to substitute a different activity or time.

This is where the wrapper’s local arrangements become operational. It should explain who can approve an alternative and how the unresolved plan is communicated. A packet that only describes the successful route leaves the receiving host to invent these decisions under pressure. The failure route deserves the same clarity as the happy path.

Contents · Previous section · Next section

SECTION 7 OF 24

7. Compensation is a new operation

Back to contents

If the plan is abandoned, releasing the held kit may compensate for the earlier reservation. The release is a new request. It has its own identity, conditions and outcome. It can fail or remain unanswered. Calling it undo can be misleading if that word suggests the earlier event never happened.

AWS’s guidance on saga orchestration describes coordinating sequences of local transactions and compensating actions, with additional complexity around retries, isolation, observation and coordinator failure. That is a useful technical comparison for this proposed workflow. We are not deploying a saga service or claiming its distributed properties for the simple lab.

The distinction is especially important once actions reach the physical world. A reservation that has not been used may be releasable. A kit that has already left the room requires a different process. A promise made to a learner may need an explanation even after the records are corrected. Compensation addresses consequences under current conditions; it does not erase history.

Contents · Previous section · Next section

SECTION 8 OF 24

8. A compensating request can remain unresolved

Back to contents

Suppose the coordinator asks the kit owner to release a reservation and receives no acknowledgement. It cannot honestly report that the resource is free merely because it sent the request. The plan should preserve a distinct status such as release requested, outcome unknown. That wording identifies both the attempted remedy and the remaining uncertainty.

A successor receiving the plan can then investigate the correct operation. Without this record, the successor may retry the original reservation, assume the kit is available or start a second plan that conflicts with the first. A good handover includes unfinished repair work, not just completed successes.

The fictional exercise can make this visible by withholding a release-response card. The participants must describe what they know and what they do not know. The goal is not to make the workflow look smooth. It is to teach the discipline of keeping uncertain outcomes separate from settled facts until there is evidence to resolve them.

Contents · Previous section · Next section

SECTION 9 OF 24

9. Use a small state vocabulary with clear meanings

Back to contents

The proposed coordination record can distinguish planned, waiting, ready, declined, release pending, closed and unresolved. These are suggested meanings for this example, not a universal standard. A host should use only the states it can define and support. More labels do not improve a workflow if participants cannot tell when each applies.

StatusMeaningPermitted claim
WaitingAt least one required result is pendingThe plan is being prepared
ReadyRequired commitments are confirmed under the contractThe recorded combination is held
Release pendingA compensating request lacks a settled outcomeRelease has been requested
UnresolvedAvailable evidence cannot establish the relevant stateA responsible owner must investigate
ClosedThe declared closure conditions are metNo further action is owed under this plan record
Proposed plan statuses for a multi-owner session

Even ready is bounded. It says the recorded combination is held; it does not establish that a real lesson has been delivered successfully. The next stage may require physical checks and participant confirmation. Each state should permit a specific statement, rather than invite the reader to infer more than the workflow can support.

Contents · Previous section · Next section

SECTION 10 OF 24

10. Keep an obligation record beside the event record

Back to contents

An event record says what happened or what was observed. An obligation record says what remains to be done, by whom and under which conditions. The two are related but different. A successful kit reservation is an event. Obtaining a slot or releasing the kit is subsequent work that may still be owed.

For Plan P, a compact record could identify the last confirmed local outcomes, unanswered requests, next responsible owner and the condition for declaring completion. It should also record whether the requester has accepted an alternative. This gives a successor a way to continue the purpose instead of merely reading a chronological list.

The proposed record is a design aid. It is not a legally authoritative register of obligations, and the example does not create real commitments. Its educational value is to distinguish remembering activity from managing unfinished work. Civilisational continuity often depends on that distinction: a successor must inherit what remains to be carried, not just what the predecessor already did.

Contents · Previous section · Next section

SECTION 11 OF 24

11. Three traces can end with the same visible message

Back to contents

Consider three ways a coordinator might display unavailable. In the first, the slot owner clearly rejected the request before any commitment. In the second, the slot owner accepted it but the acknowledgement was lost. In the third, the slot was confirmed and later withdrawn through a separate process. The same short message can conceal different histories.

The user interface may need to remain concise, but the operating record should preserve enough detail for recovery. Otherwise a successor cannot determine whether a new request is safe, whether a resource is still held or whether a participant needs an explanation about a changed commitment. Simplifying the display should not destroy the distinction in the underlying record.

A classroom group can compare these three traces and propose suitable next actions. The exercise should ask what evidence would resolve each case. Guessing the most convenient history is not a recovery method. The point is to connect a response to the specific uncertainty, using the smallest additional observation capable of deciding the next step.

Contents · Previous section · Next section

SECTION 12 OF 24

12. Order the steps according to their consequences

Back to contents

A coordinator must choose an order when operations cannot be committed together. One proposal is to secure the scarcer resource first. Another is to perform the more easily reversible action first. Those choices can conflict. The appropriate order depends on the service’s conditions and the costs of partial completion.

The fictional session makes the reasoning visible. If kits are plentiful but slots are scarce, holding a kit before asking for a slot may create unnecessary work. If a slot reservation is costly to release, the host may prefer to establish kit suitability first. The choice needs evidence about the actual resources and the permitted alternatives.

This chapter does not prescribe a universal ordering rule. It asks the host to state why a chosen order is appropriate and what happens when an intermediate step fails. A portable mechanism should carry that reasoning so a receiver can reconsider it when local scarcity, reversibility or authority differs from the original setting.

Contents · Previous section · Next section

SECTION 13 OF 24

13. Time limits require an owner and a clock

Back to contents

A held resource may need an expiry rule to avoid remaining unavailable indefinitely. Adding a deadline seems simple until different components disagree about time or about whether a delayed response arrived before expiry. The workflow must specify whose clock and record determine the outcome.

The current Composition Lab has no clock or expiry. Its slot labels are abstract tokens, and its revision counter orders accepted changes rather than measuring elapsed time. A receiving reader should not infer scheduling capability from the word slot. A real timetable would need additional representations and checks.

For the proposed multi-owner process, a deadline can be discussed as a contract field that still needs implementation. What expires: the local reservation, the overall plan or the authority to keep retrying? Who communicates expiry to the other owner? Can a late acknowledgement revive a closed plan? These questions belong in the design before the host tells participants that a reservation automatically manages itself.

Contents · Previous section · Next section

SECTION 14 OF 24

14. A handover must carry unfinished plans

Back to contents

Suppose the original coordinator leaves while several plans are waiting. A successor needs the pending requests, their identities, the last confirmed outcomes and the rules for further action. A list of completed reservations is insufficient because it omits the very work that requires attention.

The handover should also distinguish a stopped attempt from a deliberately closed plan. If a former coordinator stopped waiting for a reply, that does not necessarily close the receiving component’s commitment. The successor may need to query the owner or reconcile records before starting new activity that uses the same resources.

This extends the connected book’s question about what survives when founders leave. The inheritance includes the state of unfinished coordination, not just the institution’s purpose and stable rules. A portable packet that explains how to begin but says nothing about resuming interrupted work has transferred only part of the operating capability.

Contents · Previous section · Next section

SECTION 15 OF 24

15. Distributed ownership need not mean absent responsibility

Back to contents

Several organisations can own different steps while still maintaining an intelligible route for a participant seeking help. The participant should not have to reconstruct the entire architecture to discover why a promised session is unresolved. A host can identify a responsible contact for the overall plan while preserving each component owner’s authority over its own records.

This arrangement requires work. The coordinating role needs access to appropriate evidence and a mandate to ask for clarification or propose a remedy. It should not be assigned responsibility for outcomes while being denied any way to investigate them. The wrapper needs to state the actual relationship.

In the teaching exercise, one learner can hold the overall plan record while others own the kit and slot records. The group then examines what information the coordinator needs and which decisions remain with the owners. This is a practical way to discuss accountability without assuming that every useful system must have one all-powerful controller.

Contents · Previous section · Next section

SECTION 16 OF 24

16. AI can coordinate messages without owning the purpose

Back to contents

An AI assistant could help track which responses are missing, prepare a status explanation or suggest a sequence of next steps. Those contributions do not automatically authorise it to change the requested activity, select a different participant or make a new commitment. The mandate must state which choices it may make.

A useful assignment asks learners to give the assistant a partial trace and inspect its proposed continuation. Does it infer failure from silence? Does it claim a release completed when only a request was sent? Does it substitute a new slot without checking whether the requester accepts it? These are concrete reasoning errors to investigate.

The exercise develops the user’s broader coexistence question through a bounded workflow. Human judgement and machine assistance can contribute different functions, with evidence of their actual roles. This is not a demonstration that a language model trained on human work will autonomously form a self-renewing civilisation. That much larger claim would need its own definitions and evidence.

Contents · Previous section · Next section

SECTION 17 OF 24

17. Monetary settlement is another bounded step

Back to contents

A later service might include payment. That would add another operation and another set of conditions. A payment record would not establish that the kit and slot were suitable or that the educational activity occurred. The coordination contract would need to explain when payment is requested and what happens if another part of the plan fails.

This is where the Bitcoin lane can meet other lanes without becoming the whole road. If Bitcoin is used for a particular monetary transfer, the interpretation of that transfer remains bounded by what it records. The surrounding service still needs its own account of delivery, remedies, authority and communication. None of those capabilities appears automatically inside a monetary receipt.

The present examples use no money and make no transactions. Their purpose is to expose the interfaces a future design would need to examine. The original civilisational-crossroads thought becomes more precise when each lane contributes a defined capability and the connections between them carry explicit responsibilities.

Contents · Previous section · Next section

SECTION 18 OF 24

18. Preserve local autonomy during interruption

Back to contents

A host may lose contact with another component while still needing to serve people locally. It should know which actions remain permitted and which require confirmation that is currently unavailable. A local copy of a shared schedule does not necessarily authorise new commitments against resources another owner may have changed.

One possible design reserves a bounded local allocation in advance. Another permits preparation but defers commitment. A third pauses the affected service while allowing unrelated work to continue. These are alternative contracts with different costs; none is established by the simple Composition Lab.

The transferable requirement is to make the interruption rule explicit. A successor should not have to choose between inventing authority and stopping everything. The packet can identify the scope of local action, the records to retain and the reconciliation required when contact returns. That is a specific form of portability across changes in connectivity.

Contents · Previous section · Next section

SECTION 19 OF 24

19. Test interruption at every boundary

Back to contents

A proposed workflow should be examined at the points where one step has happened and the next has not. What if the kit is held before the slot request is sent? What if the slot commits before the reply is received? What if a release is requested before the coordinator stops? Each boundary creates a different recovery problem.

A paper trial can use sealed response cards to reveal those histories gradually. Participants should state what they know before the hidden card is opened, then revise their next action when new evidence arrives. The exercise tests whether they preserve uncertainty instead of filling the gap with a convenient assumption.

No such receiving-team trial has been conducted for this chapter. The proposal is detailed enough to attempt and to report honestly. Its report should identify which interruptions were explored, what help was supplied and whether participants could explain the changed cases. A successful local software check from the earlier lab cannot substitute for that evidence.

Contents · Previous section · Next section

SECTION 20 OF 24

20. The connection needs its own packet

Back to contents

We can now describe what the connective packet would contain: the overall purpose, component contracts, request identities, completion rule, pending states, retry conditions, compensating actions, escalation route and handover record. The wrapper would connect those rules to the actual owners, tools, time limits and local authority.

This is more than an integration script. The script might send messages, but the packet must explain what those messages mean and what a successor may infer when communication is incomplete. The operating instance then produces evidence of what actually happened, including partial outcomes and repair attempts.

The next chapter follows the return journey of that evidence. A receiving host may discover a defect in the packet or encounter an event that demands correction. The system needs a way to carry that finding back, preserve its context and determine whether it changes one local record, one release or the wider framework. Portability becomes more credible when information can travel in both directions.

Contents · Previous section · Next section

SECTION 21 OF 24

21. A matching identifier is necessary but insufficient

Back to contents

A kit confirmation and a slot confirmation may both mention Plan P while referring to different versions of its request. Perhaps the intended kit changed after the first step, or the slot owner interpreted the group size differently. Matching the plan name alone does not establish that the confirmations support the same combination.

The proposed coordinator should compare the relevant contents as well as the identity. Which kit, which slot, which release of the contract and which requested purpose does each response cover? A host need not repeat every detail in every message, but it needs a reliable way to recover the referenced request. Otherwise the evidence chain can join two true statements into a false overall conclusion.

A useful changed case supplies two confirmations with the same plan identity but inconsistent resource details. The learner’s task is to withhold the ready claim and identify the mismatch. This extends the exercise beyond counting successful replies. It asks whether the evidence actually supports the particular promise being made.

Contents · Previous section · Next section

SECTION 22 OF 24

22. Define when further attempts should stop

Back to contents

A workflow that always retries can consume resources indefinitely without bringing the purpose closer to completion. A host should decide when repeated attempts require review, when the requester should be consulted and when the plan should be closed through an appropriate process. Those decisions depend on the task and the consequences of delay.

For the fictional session, a host might stop proposing alternatives once the requester has declined them, or pause when the remaining uncertainty cannot be resolved through the available status route. The record should say why. Stopping because the purpose is no longer wanted differs from stopping because the coordinator lacks evidence or authority.

These distinctions make the successor’s job easier. They can see whether to resume an interrupted investigation, seek a new decision or leave a properly closed plan alone. A portable process includes a thoughtful end condition, rather than treating persistent activity as proof that it is still serving its original purpose.

Contents · Previous section · Next section

SECTION 23 OF 24

23. Manual coordination is a legitimate receiving route

Back to contents

A small receiving host may be better served by a person checking both owners’ records than by an elaborate automated workflow. That arrangement can still benefit from explicit plan identities, pending states and correction records. The conceptual mechanism does not require software at every step.

The host should describe the real arrangement and its limits. If one person must be present to reconcile uncertain outcomes, that dependency belongs in the wrapper. If another person can take over using the records, the handover can be examined. Automation may later reduce some work, but it should be assessed against the actual manual route rather than an imaginary process with no costs or errors.

This preserves the book’s emphasis on capability. The objective is an intelligible way to make and complete appropriate commitments. Different carriers may support that function under different conditions. The receiver should be able to choose a route it can sustain and to explain what evidence that route produces.

Contents · Previous section · Next section

SECTION 24 OF 24

24. Sources and proposal boundary

Back to contents

The multi-owner coordination contract, paper interruptions and illustrative statuses are original proposals. Composition Lab 0.1.0 remains sequential and in memory; it does not implement this distributed workflow.

Contents · Previous section

Book and reading map · Previous chapter · Next chapter