Andamiro Note Design Contest Silver Medal Winner

This blog post is in response to the article The Currency Symbol-Like Letters of the Turkmen Alphabet from Chris Shares. To summarize its contents:

Wikipedia states that the 1993–1999 version of the Turkmen alphabet contained four currency symbols as letters: ⟨£⟩, ⟨$⟩, ⟨¢⟩, and ⟨¥⟩.

Its author Christian expressed doubts about it, citing Rewriting the “Nation”: Turkmen Literacy, Language, and Power, Dr. Victoria Clement’s dissertation in 2005.

The dissertation states that ⟨$ ¢⟩ was replaced with ⟨Ş ş⟩ by 1995, with other currency symbols being phased out by 1996.

Upon traveling to Turkmenistan, Christian saw instances of old ⟨Ñ ñ⟩ instead of the current ⟨Ň ň⟩, datable to 2004 and 2006.

Upon further online image-searching, Christian found banknotes and postage stamps using ⟨$⟩ and ⟨¥⟩.

He also found a Turkmen alphabet table using all four currency symbols, but concluded from its letterforms that ⟨$⟩ and ⟨¢⟩ were intended to be ⟨Ş⟩ and ⟨ç⟩ with an upside-down cedilla above, and ⟨¥⟩ was intended to be the Kazakh Cyrillic letter ⟨Ұ⟩.

▲ A printed Turkmen alphabet table showing all the weird glyphs of the original proposal. Image retrieved from Christian’s blog post.

However, Christian did not address two oddities with the original 1993 proposal. Firstly, where did the lowercase ⟨£⟩ come from? It has never been used in any Latin or Cyrillic alphabet before, that I know of. Secondly, why was uppercase ⟨ÿ⟩ was chosen to be ⟨¥⟩, or in Christian’s hypothesis, ⟨Ұ⟩? ⟨Ұ⟩ has a lowercase in Kazakh, doesn’t it?

My hypothesis is that these letters are indeed currency symbols, and that the original 1993 Turkmen alphabet was designed for compatibility with IBM Code Page 437 first, and aesthetics second. It explains the weird letter casing correspondences and changes, though just as Christian, I couldn’t find primary sources to corroborate this online.

The full Code Page 437 character set. ASCII Latin characters are in green, except for Q, V, and X, which are not in the Turkmen alphabet and are in purple. Non-ASCII letters and symbols are in orange. Other characters, including numbers and punctuation, are grayed out. Original image.

Hi, I’m Haley, a computer science graduate fluent in Japanese and English (but not at all in Turkmen), and here’s my retrocomputing-informed educated guess as to why the 1993 Turkmen alphabet was so weird.

The Turkmen National Alphabet’s goals, besides nationalism

Then-President Nyýazow, the spearhead of the Turkmen National Alphabet project, stated that the goals of the Turkmen Latin alphabet (besides symbolizing a new Turkmen culture) included access to information technology:

However, in the end, Turkmenistan pursued an alphabet “based on its own unique letters in order to mark every single [Turkmen] phoneme. (Clement 2005, p150) [...] Moreover, in every speech [President Nyýazow] gave in 1993 [about the 1993 Turkmen alphabet], he mentioned the importance of information technology as the “number two” reason for pursuing a Latin-based script; the “number one” reason was cultural pride. (p151)

Clue 1: the Mystery Glyph

But in that case, why the mystery glyph for the lowercase ⟨£⟩? Wikipedia, citing the alphabet table image, says it looked like ⟨ſ⟩ but placed lower, having a descender. That has never been a letter in any Latin alphabet that I know of, nor an “internationally recognized sign”—although it coincidentally looks like the letter for /v/ in the Shavian alphabet for English ⟨𐑝⟩.

Methinks that the lowercase ⟨£⟩ is the smoking gun that this alphabet was designed for an IBM PC or compatible. All the letters (and currency symbols) of the original 1993 Turkmen alphabet are representable in Code Page 437, the default character set for IBM PCs which supported a range of mathematical symbols, except for that lowercase ⟨£⟩. I think that lowercase ⟨£⟩ was originally “top half integral” ⟨⌠⟩. It fits the bill: it looks like ⟨ſ⟩ but descends below the baseline.

The lack of a serif at the bottom of the mystery glyph also corroborates that this is a top half integral. The integral parts are designed as extensible symbols, with ⟨⌠⟩ and ⟨⌡⟩ being the top and bottom halves of the integral symbol, and with ⟨⎮⟩ extending the integral if more height is needed. As such, they end in a bare stroke with no adornments.

However, the long S ⟨ſ⟩ is intended as its own letter, so in serif fonts, it ends in a serif as does ⟨f⟩. (A serif is a decorative line, at the end of the strokes of a letter.) When we look at the printed alphabet table from before, we can see that the mystery glyph has no serif at the end, indicating that it’s not a long S, but a top half integral.

Comparison of “long S” ⟨ſ⟩ and “top half integral” ⟨⌠⟩ in the font Century Schoolbook, demonstrating that the bottom of the “long S” has a serif, but not the bottom of the “top half integral”. Original image.

The typesetter either must have already had a sort (or a phototypesetting glyph, given the era and presence of the Cyrillic cursive equivalent) for the top half integral; or lacking one, must have made one, looking at a printout and noticing that it didn’t have a serif on the descender.

Clue 2: the Uppercase ⟨ÿ⟩

Other casing choices also support my hypothesis. Why was capital ⟨ÿ⟩ ⟨¥⟩? If there were no problem with putting a dieresis on a lowercase ⟨y⟩, there should be no problem with putting a dieresis on a capital ⟨Y⟩, right? And if the alphabet is supposed to borrow the Kazakh capital letter ⟨Ұ⟩, why didn’t it borrow its lowercase ⟨ұ⟩? Was it just Nýyazow’s random fancy?

No, this weirdness must have had a reason. Look at CP437 again, and notice that it contains the lowercase ⟨ÿ⟩, but it lacks a capital ⟨Ÿ⟩. The team who designed CP437 (and ISO 8859-1, and Unicode) included the lowercase ⟨ÿ⟩ for French, but they left out the capital ⟨Ÿ⟩ because it only occurred in all-caps text. But the team had included the currency symbol ⟨¥⟩, anticipating dealing with Japanese companies using the Japanese yen, and with Chinese companies using the Chinese yuan.

The full IBM Code Page 437 character set. You can find ⟨ÿ⟩ below ⟨x⟩, but the capital ⟨Ÿ⟩ is nowhere to be found.

Since ⟨ÿ⟩ was the only ⟨y⟩-like character with diacritics in CP437, and it did not have an uppercase version in the code page, the designers of the alphabet were forced to use ⟨¥⟩ as its uppercase equivalent.

Impetus for Changes

With the spread of other text encodings, including ISO-8859-2, its IBM implementation Code Page 912, and eventually Unicode, these idiosyncrasies stopped having a practical purpose. I think that is why the weirdnesses were gradually ironed out. ⟨Ş ş⟩ for Turkish and ⟨Ý ý Ž ž⟩ for Czech were supported in CP912, so as soon as font support rolled out for them, Turkmen started to use them instead. Same with ⟨Ň ň⟩ for Czech instead of ⟨Ñ ñ⟩ for Spanish, though as Christian notes, this transition still isn’t complete.

All characters in Code Page 912. Original image, using the font recreation by VileR.

While this code page also contained the dollar/escudo ⟨$⟩, cent ⟨¢⟩, and yen/yuan ⟨¥⟩ signs, it left out the libra/pound ⟨£⟩, the top half integral ⟨⌠⟩, eñe/N with tilde ⟨Ñ ñ⟩, or y with dieresis ⟨ÿ⟩. I hypothesize that targeting this code page instead of CP437 further hastened the move away from the 1993 original proposal towards the current-day alphabet.

The most probable reason why some signs and inscriptions still used ⟨ñ⟩ with tilde instead of ⟨ň⟩ with háček after the 1995 revision(s) is lack of font support, followed by Unicode making both ⟨ñ⟩ and ⟨ň⟩ available at the same time, and the designer or user’s ignorance/apathy since the two diacritics look so similar, especially at small sizes.

One oddity in this timeline is that ISO-8859-2, the parent standard of CP912, was first published in 1989. Even though I can’t find direct confirmation, github.com/unicode-org/icu-data calls Code Page 912 “ibm-912_P100-1995”, so I’m pretty sure that CP912 was first published in 1995, the same year that the weirdnesses of the Turkmen alphabet started being ironed out.

Discrepancies in the Printed Alphabet Table

I think the differences in letterform that Christian notes in his article can be explained as aesthetic or practical choices, and because—and this is purely my speculation—the typesetter that printed it had suitable Cyrillic letters lying around.

The Serifless Lowercase ⟨£⟩

As stated above, the lack of a serif at the bottom of lowercase ⟨£⟩ suggests that the glyph was originally intended for “top half integral” rather than “long S”.

The “Double-Cedilla” Hypothesis is Unlikely

Christian says that lowercase ⟨$⟩ is “a ⟨ç⟩ with another cedilla on the top”, but I find that unlikely. The “double cedilla” form of ⟨$⟩ and ⟨¢⟩ can be chalked up to a stylistic preference to make them look like letters consistent with ⟨Ç ç⟩ instead of currency symbols. After all, many fonts, including the IBM PC font, have the bars missing in the middle as well.

Besides, the Bashkir, Chuvash, and Enets alphabets have a Cyrillic letter called “the”, better known as “es (⟨С с⟩) with descender”. Unlike in Latin C with cedilla ⟨Ç ç⟩, where the stroke below must at least curl towards the left, the stroke at the bottom of es with descender ⟨Ҫ ҫ⟩ could go straight down, or even curl to the right like the ogonek in Polish and Lithuanian (⟨ą ę į ų⟩). I hypothesize that the letter table uses the glyphs originally intended for Cyrillic ⟨Ҫ ҫ⟩ instead of Latin ⟨Ç ç⟩, and uses a newly-struck metal type for ⟨$⟩ and ⟨¢⟩ that fits the design of the Cyrillic ⟨Ҫ ҫ⟩.

The Kazakh-like Yen/Yuan Sign

And speaking of Cyrillic letters used in minority languages of the former USSR, since the typesetter had a Cyrillic Straight U with Stroke ⟨Ұ⟩ lying around for Kazakh text, they used it instead of ⟨¥⟩.

The Reverse-Engineered Alphabet Table

I don’t know if any “text mode” software was produced that supported the Turkmen alphabet, but here is a mock-up of what the alphabet may have looked like on a contemporary computer. Here is a table of the letters of the Turkmen alphabet using the default VGA font. Below it is Article 1 of the UNHRC translated into Turkmen, which happens to contain the rare letter ⟨ž⟩.

The original 1993-1995 proposal, represented as VGA Text Mode Art in Code Page 437.

The current version (in my timeline, proposed in 1995) represented as VGA Text Mode Art in Code Page 912.

Animated GIF comparing the two versions. The affected letters blink 4 times before switching over to the other standard.

Conclusion

My hypothesis is that the original 1993 proposal for the Turkmen National Alphabet does indeed contain four currency symbols (⟨£⟩, ⟨$⟩, ⟨¢⟩, and ⟨¥⟩), as well as one mathematical symbol ⟨⌠⟩, and it is specifically a compromise caused by the limitations of the default text encoding of the IBM PC. And as computers became more capable of representing a larger number of distinct letters, those limitations were lifted, and in 1995 with the release of IBM Code Page 912, those compromises were replaced with characters that make more sense (⟨Ž⟩ ⟨Ş⟩ ⟨ş⟩ ⟨Ý⟩ ⟨ž⟩).

Again, I can’t find primary sources for this hypothesis, nor can I even understand Turkmen. But at least, the timeline makes sense to me, and explains the weirdnesses of the 1993 Turkmen alphabet. What do you think?

Update: Daniel Bunčić

I’ve discovered that at least one serious academic has stated that Code Page 437 compatibility was a design constraint of the 1993 Turkmen alphabet before me, almost a decade ago! His name is Dr. Daniel Bunčić, and he is a professor of Slavic linguistics at the University of Cologne. (Turkmen is not Slavic but Turkic, but since Turkmen wanted a new alphabet to distance itself from its Soviet past, I think he counts as a relevant academic authority.)

Here’s an extract of his paper Factors Influencing the Success and Failure of Writing Reforms (2017), page 35, submitted to Studi Slavistici XIV (2017). (Bold emphasis is mine.) He also cites Dr. Victoria Clement like Christian did. Unfortunately, he doesn’t cite a primary source stating that Nyýazov targeted Code Page 437. Maybe I—or better, a serious academic—could inquire Dr. Clement whether she has such a direct source. (The full paper is available on Open Access, licensed CC BY-NC-ND 4.0)