Repository path: workshop/translations/maqamat-hulwan/R50-v1/translation.md · rendered 2026-09-09

«المقامة الحلوانية» — al-Ḥarīrī's second Assembly, whole, R50 v1

Source. أبو محمد القاسم بن علي الحريري (Basra, 1054–1122), «المقامات», المقامة الثانية، الحلوانية. Copy-text, provenance and its single-witness declaration: ../collation.md. Stored source: ../source-arwikisource-maqama2.txt. The whole maqāma — 140 prose cola and 17 verse lines, 709 Arabic tokens.

Why this maqāma and why now. R50 — Preston 1850's own printed remedy for Arabic rhymed prose, read as a rule — has existed on exactly one text, «المقامة الصنعانية», where it was made. A regime that has been executed once on the text it was designed against is not yet a regime; it is an instance. ARM-balanced-period needs the rule executed on material it was not built for, and needs it on a text where the man who wrote the rule has printed his own English, so that the rule as executed and the rule as its author executed it can be measured side by side on the same Arabic. Preston's Makamah of Holouan and Chenery's Second Assembly both exist and are both free.

Regime. R50 the balanced period — ../../regimes/R50-balanced-periods.md, unchanged. Prose stays prose; the sense outranks the form; the clause-ends do not chime; the two members of a pair are brought as near each other in syllable count as the sense allows; no prose colon exceeds fourteen words; verse goes into rhymed English verse.

Translated by the lead agent as a labelled subject, in session, at $0 — not ledgered (charter §3, A4). No API call produced any part of this rendering. The lead does not judge it.

Contamination: measured clean, and measured before either comparator was opened

The two published hands were located in the Internet Archive OCR by running-header line number alone and sliced to files by a script that printed only line and word counts. Neither file was opened, read, or quoted before this rendering was finished and frozen. They were then passed to tools/dependence_check.py as paths and only its statistics were read.

The lead's overlap with each published hand is smaller than the two published hands' overlap with each other — 1 and 5 shared 7-grams against their 14, and runs of 7 and 8 against their 10. Under --unicode tokenisation every figure is identical except the lead's self-overlap, which falls from 6 to 5. clean on the tool's verdict, and contamination: none on this page's front matter.

The self-overlap is the one number worth remarking, because note (bhb) records that this hand normally matches itself at up to 37 contiguous tokens — above its record against any published human translation. Here it matches itself at 6 tokens, the lowest figure the note has yet seen, and the reason is visible: R50 on «الصنعانية» and R50 here are the same rule applied to two different stories, not two renderings of one text. The self-overlap warning is about re-rendering, and this is not one.

The rendering

The prose, one record per colon, is ../cola-R50.tsv, frozen with this page. The verse is below. Prose and verse together, as one plain text, are in the experiment's raw directory.

Verse

b001–b002 (after p041) —

By him I cleared my cares away, and I could see my age turn open-faced, and bright as bright could be. His nearness I called kin, his house a rich domain, his look I drank like water, and his face was rain.

b003–b004 (after p054) —

Since he went far, no man I met has given me cheer, nor has the one who led me toward him drawn me near. No equal to his worth has shown himself to me, nor any man of parts possessed such parts as he.

b005 (after p073), the line of al-Buḥturī the young man admires —

She seems to smile upon a row of ordered pearls, or else on hail, or camomile as it unfurls.

b006–b007 (after p082), Abū Zayd's answer to it —

My soul be ransom for a mouth whose smile is rare, graced with a coolness of the teeth beyond compare. It parts upon moist pearls and hail, and then makes room for camomile, for palm-spathe, and for water's bloom.

b008 (after p112), the challenge —

Then she rained pearls from a narcissus, watered rose, and bit upon the jujube with the hail she throws.

b009–b010 (after p114), the answer to the challenge —

I asked her at her visit to lay by the red veil that she wore, and give my ear sweet news instead. A twilight she drew back that had a moon concealed, and from a fragrant ring the pearls were all revealed.

b011–b012 (after p121), the two further verses —

She came, the day the parting was in earnest, dressed in black, and bit the fingers of a man hard pressed. Then night shone over morning, on a single stem, and crystal wore the marks her pearls had set in them.

b013–b017 (after p138), Abū Zayd on time —

The stroke of troubles whitens hair, and time turns men now here, now there. If it submits to one today, tomorrow it will have its way. Then trust no flashing of its light: its lightning is a lie, though bright. Endure, when it has set the pack of troubles on you at your back. There is no shame upon the gold that in the fire is turned and rolled.

Compliance with the regime, measured

Computed by ../../experiments/E-20260826-balanced-period/metrics.py over the 140 prose cola and their 139 consecutive pairs, before the design of any study using this text.

For comparison, R50 on «المقامة الصنعانية» came in at mean |Δ| 2.174, 68.1% within 2, 0 STRICT and 5 NEAR, with three cola over the fourteen-word ceiling. The rule executes tighter on the second text than on the first, which is what a rule executed twice ought to do and is not evidence about anything else.

The six UNKNOWN pairs are all transliterated proper names — Ḥulwān, Abū 'Ubāda, Sarūj — which have no CMUdict pronunciation and which the tool therefore declines to grade. They are reported rather than counted either way.

The declared exceptions

• p001, the frame formula, is the single worst-balanced pair in the maqāma at 20 syllables against p002's 10. Al-Ḥārith ibn Hammām related is nine syllables of English apparatus sitting inside the copy-text's first colon; measured on the members alone the pair is 11 against 10. The whole-colon figure is the declared one and is what the table reports. This is the same artefact, at the same colon, as R50 on «الصنعانية», and it is left in for comparability rather than hidden.
• Five of the eight worst pairs are the verse-introducing formulae — p081/p082, p111/p112, p112/p113, p121/p122, p105/p106. «وأنْشَد» and «فابتَدَر» are one and two Arabic words against neighbours of six and seven. The source is lopsided at exactly these points, and rule 2 forbids padding the short member to match.
• p045/p046 and p074/p075 are lopsided in the Arabic too (6 words against 2; 4 against 2).
• p060 is the longest colon at 13 words and clears the ceiling; the copy-text's colon there carries a relative clause («التي هيَ مُنتَدى المتأدّبينَ») that cannot be shed without dropping the sense.

Translator's log — frozen 2026-08-26, before any evaluation of this rendering was designed

Decisions and alternatives. Not quality claims. Written before the comparators were opened.

1. The hardest thing in this maqāma is not the rhyme; it is that the plot is a literary argument. «الصنعانية» is a beggar's performance. This one is a scene in a library, and its whole middle is two men disputing which of two Arabic couplets better describes a mouth — pearls, hail, camomile, palm-spathe, bubbles. The prose around that dispute is ordinary narrative; the couplets carry the point. Under R50 the prose gives up its sound and the verse keeps its rhyme, so the maqāma's centre of gravity moves onto the verse in a way it does not in the Arabic, where the prose chimes as insistently as the verse does. That is a consequence of the regime, and it is worth recording that the regime does not distribute its losses evenly across a text: it costs most where the prose is doing the most.

2. Where the rule's three clauses collided, and what gave way. Rule 2 (the sense) outranks 3–6, and it was invoked four times.

3. Four places where the sense was genuinely uncertain and the choice is recorded so it can be disputed.

• p025 «ومُدراةٍ ودِرايَةٍ» — مِدْراة is a comb; دِراية is discernment. The pair is a jinās (near-identical consonants, different sense) and the joke is that grooming and knowledge sound alike. Rendered care for his hair and a quick understanding: the sense of both, none of the jinās. An alternative that keeps a little of the play — well-combed and well-read — was rejected because well-read is not دراية and because the four-syllable line would have wrecked the balance with p024.
• p047 «عدَمُ العُراقِ» — عُراق is a bone with meat still on it; the phrase is the absence of a meaty bone, i.e. destitution, and it puns against العِراق Iraq in the next colon. Rendered literally as want of a meaty bone rather than glossed to hunger, because the literal image is what sets up the pun, and the pun itself is not recoverable.
• p090 «لنَجيّكُمْ مُذُ اليوْمَ» — نَجِيّ is the person one confides in. Abū Zayd is claiming the verse as his own by a circumlocution: the man who confides in you today. A flatter the man speaking to you now was rejected: the evasion is the character.
• p131 «وقدْ أقْمَرَ ليلُه الدّجُوجيُّ» — literally his pitch-black night has become moonlit, i.e. his black hair has gone white. Kept as the image (whose pitch-black night had turned to moonlight) rather than glossed, because p136 says what has turned your beard to grey four cola later and the gloss would have made the narrator's question redundant.

4. What the balance rule actually does to English prose, as felt from inside it. Rule 4 has one effect the regime page does not name and that this rendering made obvious: it pushes the English toward the paired construction the Arabic already has, but by a different route. Where al-Ḥarīrī balances by rhyme and metre, an English hand obeying rule 4 balances by grammatical parallelism — for the excellence of his gifts ⁄ for the breadth of his recitation, he had beauty of person ⁄ he had care for his hair. Once the two members share a frame, matching their syllables is nearly free. The cheapest way to satisfy rule 4 is therefore to repeat a construction, and the cheapest way to satisfy rule 3 is to vary the word at the end of it. Those two pressures do not fight; they compose. That was not obvious before doing it, and it is the closest thing this rendering has to a craft finding. (No claim is made here about whether the result reads well: internal-judgment-only would be needed for that and none is asserted.)

5. What was given up, listed. The saj' — every colon of it, 140 times. The jinās at p025 and p047. The plural of الرفاق at p051. The Qur'anic cadence of p097 («إنّ بعْضَ الظنّ إثْمٌ», sura 49:12), which is rendered as a flat English sentence because rule 3 forbids the chime that a scriptural tag would want. Nothing was added: no fact, event, referent or evaluation that the Arabic does not carry.