The Publishing Protocol

Twenty-two steps, in order. Every published interpretation on this platform is produced by running them. Steps are not optional and are not reordered; several exist specifically to catch errors introduced by skipping an earlier one.

§17.1 Phase A — Establish the text

1. Locate the passage in a printed edition and record the edition. Not a website, not a recollection, not a quotation inside another work. If the passage cannot be located, stop and file it in the open register at Grade C or D. A very large proportion of interpretive disputes dissolve here.
2. Record variants. Where recensions differ, record every reading with its witness. Do not silently adopt the convenient one.
3. Fix the boundaries. Determine where the passage begins and ends as a unit of argument. A verse quoted out of its adhikaraṇa is a different object from the verse in place.
4. Produce the anvaya. Resolve compounds, identify cases, state grammatical relations explicitly. CF-06 shows why: a change of parse can invert a verse. Where more than one parse is available, record all of them at this stage rather than choosing.
5. Translate literally. Own translation, deliberately unpolished. Felicity comes later and is a separate artefact.

§17.2 Phase B — Establish the subject

6. Run the six marks (§8.2) on the containing section. Record the output of each mark separately, including where a mark is silent.
7. Rank conflicting marks (§8.3) by the Jaimini hierarchy. Note explicitly if the reading depends on rank five or six; such a reading is weak and must be labelled.
8. Identify the register or registers (§5.2) at which the passage is operating.
9. State the doubt in one sentence. If this cannot be done, the analysis is not yet ready; return to step 6.

§17.3 Phase C — Test the reading

10. Run S1: list the predicates the passage attaches to the disputed term.
11. Run S2: test the ordinary referent against every predicate.
12. Run S3: halt if it passes. This step is mandatory and is the most frequently skipped step in the entire protocol (failure mode M03).
13. Run S4 only on overflow — where predicates exceed the ordinary referent.
14. Run S5: seek corroboration in a source at tier T1–T4. Record Grade A on attestation, Grade B on inference.
15. Check the licensing conditions (§9.2) if any word has been taken in a secondary sense. All five, stated. If L1 cannot be established without assuming the conclusion, refuse the reading.
16. Check derivations against N1–N5 (§7.5) if any nirukta reasoning is used.

§17.4 Phase D — Test against the corpus

17. Run the conflict protocol (§10.5) against every passage known to bear on the question, including — especially — passages that tell against the reading.
18. Count the strain. How many other passages must be read non-obviously for this reading to hold? Record the number. A reading with a lower strain count beats a more congenial one with a higher count.
19. Construct the pūrvapakṣa at full strength (§8.4) — ideally from a source that actually holds it, in its own words, rather than from imagination.
20. State the defeat condition. What evidence would overturn this reading? A reading with no defeat condition is failure mode M04 and is not published.

§17.5 Phase E — Publish

21. Grade it (§0.4), applying rule C-03: a convenient conclusion is graded one level more conservatively.
22. Address it (§19), attribute it, and file the residue. Assign URNs to mūla and to each commentary node separately. Name every human author. Enter every unresolved item in the open register with both readings stated.