Repository path: workshop/regimes/R51-radif.md · rendered 2026-09-09

R51 — the radif carried whole, and the same line-ends with each half taken away

v0.1, pilot. R44's craft clauses, plus a 2×2 at the line-end.

The Persian rhyming position is […qāfiya][radif]. Two formal resources sit in one slot on every line: a word or phrase repeated identically at every rhyming position, and a chime on the word immediately before it. R44 and R47 were built to carry a chime and to move it. Neither can ask what the repetition is worth, because in the Gulistan's قطعه the repetition is incidental where it occurs at all.

A ghazal is the form where it is not incidental: the radif runs the whole poem, at the poem's own depth, and the rhyme never falls at the line-end at all.

The rule

1. These clauses of R44 carry over unchanged and outrank everything added here: - Render the sense. No fact, event, referent or evaluation added; nothing dropped. A rhyme bought with a word the Persian has not got is a refusal, not a rhyme. - Enumerate before choosing, and write the enumeration down, whether or not it yields. - Grade every rhyming position mechanically with tools/rhyme_pairs.py after the hand has written its own reading down, both standing where they differ. - The hand does not revise a rendering to move a grade. - Verse stays verse. - Freeze the log before any evaluation is designed and before any published hand is opened.
2. The rendering of record is AB. At every rhyming position of the ghazal — both hemistichs of the مطلع and the second hemistich of every subsequent bayt — the English line ends in the same word or phrase, identical at every position, and the word immediately before it rhymes with the corresponding word at every other position, to the source's own depth. This is the translation this work files as its artifact and the only one graded HELD / SHORT / REFUSED / UNREACHABLE.
3. At the same sitting, from the same Persian, the hand writes three declared variants, which differ from AB only in the line-end material and in nothing else on the page: - RH — the chime kept where it is, the repetition destroyed: the tail phrase is replaced at every position by a different, sense-defensible wording, so that no two positions end alike and the chime still falls in the pre-tail slot. - RD — the repetition kept, the chime destroyed: the bearer word is replaced at every position by a sense-defensible word that rhymes with nothing. - NN — both destroyed.
4. Every substituted word is defensible on rule 1's sense clause alone and is recorded with the reason it is acceptable. Where no acceptable substitute exists at a position, the position is declared NO-ALTERNATIVE and the poem supplies no item to the arm that needed it.
5. The variants are not translations of record and are never filed as such. They are instrument material; the artifact page prints them under a heading that says so, and no claim about how one ought to translate a ghazal is made from the fact that the hand wrote them.
6. Everything in rules 3 and 4 is written before the experiment that uses it exists, and the log is frozen and committed before the design is written.

Why the chime's position is the same in all four arms

RH does not promote the chime to the line-end. If it did, the contrast between AB and RH would confound two changes — the repetition going away and the chime moving — and this project has already measured that a chime's placement is worth something on its own (RS-20260824c-run-placement, +0.107 between two placements). Holding the tail present but varied in RH, and present and identical in AB, isolates the repetition and nothing else.

What R51 is not

It is not a recommendation that a ghazal be Englished this way, and the four arms are not four candidate translation policies to be ranked. AB exists to answer whether the shape can be hosted at all; RH, RD and NN exist so that the two halves of one device can be measured apart. The known limitation, declared here rather than discovered later: one hand writes all four arms and that hand knows what the experiment asks. The design's protection is that the substitutions are confined to the line-end and that both halves are removed in both orders, so a hand that writes worse English whenever it removes something shows up in RD and RH alike and is subtracted.