Document model: how a unit that holds units of its own type is stored, reused and checked

Question: how is a unit that can hold units of its own type stored, so that it is a tree a reader can follow, reusable in several places and checkable by a program? And what do document models and their critics say?

Files

File Role
protocol how the pool on document models, reuse and checking was searched and scored
sources the pool, counted stage by stage, and each source with its quality
frame step 1: the problem, and what markup research and practice call it
review step 2: OHCO and its critics, content models, reuse, cycles, checking
state of the art step 3: how fourteen systems treat recursion, reuse and cycles
position step 4: the answers on self-reference and graphs, and candidate rules
against the cost of strict models, and fragments reused without their context
baseline step 5: DITA maps and conref, DocBook with RELAX NG, and OHCO
bibliography every cited work, shared by the three reviews
flow.json the flow as data, drawn in the figure of sources

Claims

Claim Grade Where
Document standards put recursion in the type and keep each instance a finite tree convention (standards), descriptive review.md 1, 4
DocBook and TEI offer a capped numbered form (5 and 7 levels) beside the recursive one convention (standards) review.md 1
Recursion is common in schemas and shallow in instances: 58% of DTDs hold recursive elements; 99% of web XML documents have fewer than 8 levels measured, descriptive (secondary) review.md 5
No version of the OHCO thesis is free of counterexample supported with caveat (argument from cases) review.md 2
Where reuse is not enforced, copies are common: mean clone coverage 13.6% over 28 specifications, up to 71.6% measured, descriptive, one study review.md 3
Clone detection finds duplicated prose at precision above 85% after tailoring; recall was not measured measured, descriptive, one study review.md 5
In 2008, half of 276 respondents used single sourcing or content management; half of its users reported significant downsides; about 1 in 10 had seen a failure measured, descriptive (survey, secondary) review.md 3
Reused components lose context for translators, and the sources disagree on quality supported with caveat (integrative review of 60 articles) review.md 0; against.md
85.4% of 180,000 web XML files are well-formed; about a third of those that name a schema validate measured, descriptive (secondary) review.md 5
Every system read on cycles that allows references forbids, reports or caps a cycle, or leaves it undefined; none keeps a self-containing unit convention, descriptive count of 14 systems state-of-the-art.md
A unit with two parents is allowed only by reference, with a rule per use for links, navigation or permissions convention, descriptive position.md
Automated writing feedback improves students' writing, g = 0.55 measured (meta-analysis, secondary; students, not documentation) review.md 0
The evidence for design patterns in general is weak; none of it concerns documents measured, systematic review (secondary) review.md 0
Reuse by reference helps readers or keeps documentation consistent not found: no study in this search position.md
One containing parent, acyclic references, each reuse valid in each place convention, untested (candidate rules) position.md

The two questions this review answers

  • Self-reference. The template needs the recursive type, a topic whose children are topics. No model read needs a topic that contains itself, and standards forbid it by syntax, by a rule on references, or in the processor (position).
  • Graphs. In the standards read, the stored form is a graph of trees (files that reference files), each reading is a tree, and what crosses the hierarchy is held beside it as pointers. No study shows which view helps readers (position).

Named baseline

  • To beat: DITA 1.3 topics and maps, reuse by conref and keys, validated, with a processor that reports cycles (OASIS 2015; DITA-OT 2026b).
  • Single-document twin: DocBook 5.2 recursive section with RELAX NG and XInclude (Walsh 2019; Clark 2001; Marsh 2006).
  • Model to beat as a claim about text: OHCO (DeRose 1990), with its critique (Renear 1996).
  • Comparison: seeded structural defects, consistency after a change, reading a reused unit in a new place, cost to the writer (baseline).

General findings linked, not reviewed again