The Principia Machina/GTP/Walk/5 Keeping the pieces in step - Revision history

BFP Rauch: Edited under the disclosure rule

2026-10-13T07:01:05Z

Edited under the disclosure rule

← Older revision		Revision as of 07:01, 13 October 2026
Line 10:		Line 10:

	The answer is that every piece declares three things about itself, in its own header, where a reader and a machine can both find them. Its own version. The range of profile versions it is built to work with. And the sections of other documents it actually relies on.		The answer is that every piece declares three things about itself, in its own header, where a reader and a machine can both find them. Its own version. The range of profile versions it is built to work with. And the sections of other documents it actually relies on.

	~~The diagram shows what happens with those declarations at the moment a piece is loaded.~~

	~~{{:The Principia Machina/Diagrams/P05 Compatibility and version semantics}}~~
	~~''What each artifact declares, and what happens when the declaration does not match.'' ([[The Principia Machina/Diagrams/P05 Compatibility and version semantics\|diagram page]].)~~

	The declared range is compared against the profile that is actually in force. If it matches, the piece runs. If it does not, or if the declaration is missing altogether, the piece says so and stops. It does not run with a warning attached, because a warning is something to be read past, and the whole point is that this particular failure is invisible in the output.		The declared range is compared against the profile that is actually in force. If it matches, the piece runs. If it does not, or if the declaration is missing altogether, the piece says so and stops. It does not run with a warning attached, because a warning is something to be read past, and the whole point is that this particular failure is invisible in the output.

BFP Rauch: CO-4 rebuilt against PMP-TPM 0.13.0 (D185.1 (c); D174.1): relabel and mirror update

2026-09-27T01:01:32Z

CO-4 rebuilt against PMP-TPM 0.13.0 (D185.1 (c); D174.1): relabel and mirror update

← Older revision	Revision as of 01:01, 27 September 2026
(No difference)

BFP Rauch: 1 revision imported

2026-09-27T00:13:40Z

1 revision imported

← Older revision		Revision as of 00:13, 27 September 2026
Line 1:		Line 1:
	{{TPM map}}		{{TPM map}}
	{{DISPLAYTITLE:Keeping the pieces in step}}		{{DISPLAYTITLE:Keeping the pieces in step}}
	''The GTP walk, page 5 of 6. Written for the general reader; nothing here is a rule~~, and every~~ rule it mentions is stated by [[The Principia Machina/GTP\|the specification]], which governs wherever ~~the two~~ differ.''		''The GTP walk through, page 5 of 6. Written for the general reader; nothing here is a rule. Every rule it mentions is stated in [[The Principia Machina/GTP\|the text published under the name GTP]], documentation beside [[The Principia Machina/TPS\|the specification]], which governs wherever they differ.''

	== Keeping the pieces in step ==		== Keeping the pieces in step ==
Line 30:		Line 30:

	[[Category:The Principia Machina]]		[[Category:The Principia Machina]]

BFP Rauch: GTP section: specification, Claude instance, conformance, walk, guide, agent pages, diagrams

2026-09-18T18:00:00Z

GTP section: specification, Claude instance, conformance, walk, guide, agent pages, diagrams

← Older revision	Revision as of 18:00, 18 September 2026
(No difference)

BFP Rauch: 1 revision imported

2026-09-18T06:42:59Z

1 revision imported

New page

{{TPM map}}
{{DISPLAYTITLE:Keeping the pieces in step}}
''The GTP walk, page 5 of 6. Written for the general reader; nothing here is a rule, and every rule it mentions is stated by [[The Principia Machina/GTP|the specification]], which governs wherever the two differ.''

== Keeping the pieces in step ==

The instruction set is not one document. It is a profile, a project layer, and a handful of separate procedures, and each of them is revised on its own schedule by its own round of work. That is the sane way to maintain them, and it creates a specific hazard.

A procedure written against last year's profile can rely on a section that has since been reworded, or on a rule that has since moved. Nothing announces this. The procedure still loads, still reads sensibly, and still produces output that looks like its output always looked. It is simply now built on something that is no longer there.

The answer is that every piece declares three things about itself, in its own header, where a reader and a machine can both find them. Its own version. The range of profile versions it is built to work with. And the sections of other documents it actually relies on.

The diagram shows what happens with those declarations at the moment a piece is loaded.

{{:The Principia Machina/Diagrams/P05 Compatibility and version semantics}}
''What each artifact declares, and what happens when the declaration does not match.'' ([[The Principia Machina/Diagrams/P05 Compatibility and version semantics|diagram page]].)

The declared range is compared against the profile that is actually in force. If it matches, the piece runs. If it does not, or if the declaration is missing altogether, the piece says so and stops. It does not run with a warning attached, because a warning is something to be read past, and the whole point is that this particular failure is invisible in the output.

It is a range, with a top and a bottom, rather than a minimum. A minimum is a promise about releases that do not exist, made by someone who cannot have tried them. The range says what was actually tested and leaves the next version to state its own case.

The numbers themselves are not decoration. Three parts, each meaning something definite: one moves when something that other documents depend on has changed in a way that breaks them, one when something was added that does not break anything, and one when a mistake was fixed. Reading the numbers tells you whether you need to look.

What counts as breakage is defined per kind of document, which is subtler than it sounds. For an instruction artifact it is the cues it answers to, plus any section another document names. A section nothing else refers to can be rewritten freely; a section someone else depends on cannot, even if the rewrite is an improvement. For a body of findings it is the claims and the labels others cite.

One restriction is aimed squarely at machines. Work running with nobody watching may never claim more than a fix. Deciding that a change is larger than that is a judgement about what other people were relying on, and a judgement of that kind needs somebody present to make it.

----
Previous: [[The Principia Machina/GTP/Walk/4 What the agent will not do|What the agent will not do]]. Next: [[The Principia Machina/GTP/Walk/6 Saying what the work is for|Saying what the work is for]].

[[Category:The Principia Machina]]