Lost in TranslationGuardian (UK)
Transcribing Arabic into the Roman alphabet is fraught with difficulty. And in an age of electronic text, search engines and databases, the problem is only going to get worse.
Searching the BBC's vast website for articles about Colonel Gadafy recently, I found just three mentions of his name.
As far as the BBC is concerned, this may be three times too many, since its approved spelling of the Libyan leader's name is "Gaddafi".
Here at the Guardian, on the other hand, our policy is to call him Gadafy - something we have succeeded in doing 325 times on our website.
For good measure, we have also managed to write Gaddafi 42 times, Gadafi eight times, and Gaddafy and Qadhafi twice each.
If given a completely free choice in the matter, I would much prefer to call him Colonel Qadhdhaafiy because that spelling, besides being more faithful to the original Arabic, accurately reflects the eccentricity of the man.
Arguments about the ideal spelling of Arabic names go back at least to 1926, when TE Lawrence ("Lawrence of Arabia") sent his 130,000-word manuscript of Revolt in the Desert to be typeset.
A sharp-eyed proofreader complained that it was "full of inconsistencies in the spelling of proper names, a point which reviewers often take up".
Among other things, the proofreader noted that "Jeddah" alternated with "Jidda" throughout the book, while a man whose name began as Sherif Abd el Mayin later became el Main, el Mayein, el Muein, el Mayin and le Muyein.
Lawrence, who relished such vagaries, told his proofreader to get lost. "Arabic names," he replied, "won't go into English exactly, for their consonants are not the same as ours, and their vowels, like ours, vary from district to district.
"There are some 'scientific systems' of transliteration, helpful to people who know enough Arabic not to need helping, but a washout for the world," he continued. "I spell my names anyhow, to show what rot the systems are."
Up to a point, Lawrence was right. Inconsistent spelling can certainly be consistent with the inconsistency of Arabic pronunciation, and in the 1920s - apart from the occasional tut-tutting book reviewer - nobody would have minded very much.
Today, though, life is more complicated. Even the humble telephone directory is becoming a problem. In Lawrence's time there were few Arabs living in the London and even fewer with telephones.
But try looking them up today. An Arab known as Hassan al-Mughrabi might be listed in any one of five places: as "Al Mughrabi", "Al-Mughrabi", "El Mughrabi", "El-Mughrabi" or just plain "Mughrabi".
Confusion over the spelling of Arab names that have been transcribed into the Roman alphabet can also lead to more serious problems. Efforts by the FBI to track down Osama (Usama) bin Laden's (Ladin's) supporters have been hampered by this, and a recent libel case in Britain was complicated by the way spellings of Arab names changed from document to document.
The obvious solution is to have a standard, internationally agreed, system for converting Arabic script to the Roman alphabet, but that is easier said than done.
For a start, only eight Arabic consonants - B, F, K, L, M, N, R, and Z - have an indisputably equivalent letter in the Roman alphabet. Arabic also has two distinct consonants that sound like S, and the same applies to D, H and T.
In addition, there are some glottal sounds in Arabic that have no obvious Roman equivalent. This leaves plenty of scope for scholarly debate, with the result that there are now many supposedly international standards.
One of the earliest was that adopted by the International Convention of Orientalist Scholars in 1936. Another was agreed in 1971 at a conference of Arab experts in Beirut and accepted - at least in theory - by the countries of the Arab League.
Besides these, there is ISO 233, DIN 31635 and even a British standard, BS 4280, which people are actively discouraged from finding out about or using. The copyright of BS 4280 is closely guarded by the British Standards Institute which charges £28 ($39) for an eight-page booklet explaining the system.
Slightly more successfully, the US Library of Congress and the American Library Association have issued "Romanisation tables" covering more than 150 languages and dialects (including Arabic) that are written in non-Roman scripts.
The system, originally devised for cataloguing books, has found its way into wider academic use, and detailed explanations of how it works can be found on several websites.
Meanwhile UNGEGN (the United Nations group of experts on geographical names) has spent the past 30 years trying to standardise the spelling of place names on maps and similar products with varying degrees of success.
There are several reasons why all these Romanisation systems have failed to catch on. One is that they are too clever for general use, requiring dots, lines and other marks on certain letters which are impossible to achieve with a typewriter or ordinary word-processor - quite apart from baffling the average reader.
Another factor is that in some Arab countries - especially in North Africa - French is more widely spoken than English. In these, the preference is to Romanise Arab names in ways that correspond more closely to French pronunciation.
So, for example, the Arab surname often spelled as Shaheen becomes Chahine in French-influenced countries.
There are also a lot of vaguely phonetic spellings - such as "Koran" and "Mecca" - that bear little relation to the Arabic spelling but entered popular usage many years ago and are now difficult to eradicate.
An internet search with Google shows that "Mecca" is used almost six times more often than "Makkah" (the more accurate spelling that Muslims generally prefer). In a similar search "Quran" scores 44%, "Koran" 37%, and "Qur'an" 19%.
Spellings of the most common Arab name found by Google were: Muhammad 41%, Mohammed 32%, Mohamed 25%, Mahomet 3%.
Strictly speaking, there's no reason why E or O should ever appear in an Arab name, since Arabic has only three vowels - A, I and U.
Many erroneous spellings have become so entrenched and familiar to us that more accurate versions leave us mystified on the rare occasions that we encounter them.
Last week I spotted a report in one of Britain's least-read daily newspapers (News Line, published by the Workers' Revolutionary Party) which mentioned a Palestinian politician called Sa'ib Urayqat. After puzzling over this for a moment, I realised it was talking about the man usually known as Saeb Erekat.
The spelling adopted by the Workers' Revolutionary Party is certainly more logical but, as the party ought to have discovered by now, having logic on your side does not necessarily attract followers.
Underlying the confusion is a tussle between the spoken and written word. Early travellers to the Middle East usually wrote down words as they heard them (or thought they heard them) in ways that readers back home would find easy to pronounce.
This phonetic approach, which is the one Lawrence favoured, naturally takes account of any local quirks in pronunciation.
Some of the differences are significant - for example, J is pronounced as G in Egypt, while Q is pronounced as G in the Gulf - and the result is that a single Arabic word, spoken by a Moroccan, an Egyptian and a Saudi could easily appear as three different words if written phonetically in the Roman alphabet.
On the other hand, transcriptions based on the written form of Arab words (which is standard throughout the region) are less likely to be ambiguous in meaning, even if they miss out on local colour.
But the conflict between the written and spoken word is never likely to be fully resolved, especially in North Africa, where pronunciation differs greatly from classical written Arabic. President Bouteflika of Algeria is one person whose name would become unrecognisable if transcribed classically.
Most of the ideas about how to transcribe (or how not to transcribe) Arabic words into the Roman alphabet were developed long before computers became an everyday tool - and now there are those who say the whole approach has been wrong.
So far, we have only been concerned with one-way transcription of Arab words into the Roman alphabet. But in the future, with multilingual databases and so forth, we shall need to use the two alphabets interchangeably.
So, for example, electronic text in Arabic would have to be converted automatically into the Roman alphabet and then back again into flawless Arabic.
Research in this area has been led by the Xerox company, and there's an interesting but technical discussion of the issues by Kenneth Beesley which can be found on the internet (see links below) [note: when I mirrored this article from www.guardian.co.uk/elsewhere/journalist/story/0,7792,730805,00.html, there were no links provided. -J.W., 2010].
At a more practical level, there's an ingenious system called the Buckwalter Transliteration. This was devised by Tim Buckwalter, a lexicographer, for sending Arabic-language emails from computers where the keyboard and display system allow only the Roman alphabet to be used.
These developments hold out some fascinating prospects for the future. But I fear it will still be some time before the world agrees on how to spell Gadafy.