Monday, October 12, 2015

The secret life of Bambara Arial

Quite unexpectedly, I heard yesterday the names of two old 8-bit Malian computer fonts - Bambara Arial and Bambara Times - from two different people. Matt Heberger, who is coordinating work on the forthcoming Bambara translation of Where There is No Doctor mentioned that one of the translators provided him with copies of these fonts (as .ttf files), and Sam Samake, an old friend and former Peace Corps/Mali staff member, asked how I got "Bambara Arial" to work on the internet.

It wasn't supposed to be this way once Unicode rendered such special fonts unnecessary.

The history

In the 1990s, a joint project of the Malian Ministry of Education and the French Agence de Coopération Culturelle et Technique (ACCT; precursor to the Organisation Internationale de la Francophonie), produced two modified/hacked versions of the Arial and Times fonts, replacing characters not used in the orthographies of Malian national languages with characters not present in those original fonts or other commercially available fonts of that era. Thus for example, the "q" was replaced with "ɛ"* (type "q" get "ɛ" instead).

Table from Enguehard & Mbodj 2004.* Caractère affiché is what you get;
Caractère initial is original character & the key you tap to get the new one.
This approach for adapting fonts for writing systems not supported by the pre-Unicode standards was fairly common 1980s and 1990s, including in a number of African countries like Mali. Fonts like these achieved what they were primarily intended for - being able to compose and print documents with the extended alphabets. One could also share digital copies of documents, but those could be properly read only with the same fonts. Different modified fonts - and Mali had several - were mtually incompatible. That was the whole reason, of course, for Unicode, which also makes it possible to share documents in any alphabet across the internet on any browser, wordprocessor, etc.

But while Unicode became the international standard, evidently at least some people in Mali kept using Bambara Arial and perhaps other similar "special fonts." In 2005, USAID-funded "Community Learning and Information Centers" (CLICs) relied on these fonts for anything done in Malian national languages (apparently not that often). It may be that technicians in these telecenters did not have Unicode explained to them in their project training or prior study of computers.

The word about obsolescence of 8-bit fonts like Bambara Arial may not have gotten too far, or maybe the notion of a need for a "special font" to process text in languages like Bambara just was too ingrained. At this point I'm just wondering how after almost 2 decades, these old fonts are still in circulation and conversation. Just two years ago, there were these references to Bambara Arial online (thanks to a Google search):
  • N'oublie pas de les ecrire en vrai Bambara ''Bambara Arial''Dans Microsoft Word (Facebook, 2013-1-11)
  • I would like to arrange for volunteer translators for Bambara. How can I access fonts for the Bambara alphabet (Bambara Arial for example)? (Google code Khan Academy issue tracker, 2013-3-23)


More to it?

Adapted from an image by Denis Jacquerye.

There's a twist to the story though. It seems that in one respect, these old fonts follow Malian orthographies better than the Unicode fonts. The letter ɲ has two forms of upper case, one like the lower case letter but bigger ("n-form") and the other like the capital N with a tail on the left leg ("N-form") on the left and right sides respectively of the image on the right. The "n-form" is most used in Mali and apparently in the hacked 8-bit "special fonts" like Bambara Arial; most Unicode fonts use the "N-form" (thanks to Matt Heberger for his observations on that.). I doubt that this alone could account for the persistence of the old fonts, but it might be a factor.

Maybe a new life could be given to the old fonts that people are still using by reencoding them to Unicode and releasing them under recognizable names.

* Two articles in French mention modified 8-bit fonts used in Mali, showing which different characters were changed to extended Latin characters:  
Chantal Enguehard,et Chérif Mbodj. "Correcteurs orthographiques pour les langues africaines." Bulag 29, 2004, pp. 51-68.
Chantal Enguehard, et Soumana Kané. "Langues africaines et communication électronique : développement de correcteurs orthographiques." Agence universitaire de la Francophonie. Actes des Premières Journées scientifiques communes des réseaux de chercheurs concernant la langue, 31mai-1 juin 2004, Ouagadougou, Burkina Faso. pp.59-75.


Andj said...

The glyph variation by itself isn't a show stopper since some fonts will have the required glyph and it would be possible to add to opensource fonts.

One thing that may be lacking are tools to convert from the legacy encodings to unicode.

Additionally for people who know how to type with those legacy fonts may also require a keyboard tbat matches the way they are used to typing.

Don said...

Thanks Andrew. You're right that one glyph itself wouldn't necessarily make the decision, but I'm thinking that as part of a familiar package (and alternative glyphs in most Unicode fonts) it discourages change to Unicode. The keyboard aspect is probably biggest for people used to the Bambara Arial system.

Wrt conversion utilities, Bill Poser used to have a page explaining how to use a utility that the defunct FontForge had. Was assuming since the concept is not new, there should be currently available tools for conversion to Unicode.

Then it's back to the keyboard issue - perhaps have to associate a re-encoded Bambara Arial with a Keyman keyboard layout or something. All in the absence afaik of any proposed keyboard layout standard for Mali and the region. These discussions of course go back a decade and a half (as you of course know well) - and could be one topic for any conference marking 5 decades since the first efforts to "harmonize" the orthographies in the region.)

Andj said...

I assume yo meant Bill's page at

With respect to conversion I wasn't thinking in terms of conversion of fonts, but of data. I.e.

c9 said...

In Japan/Korea, as they use their local currency symbol to replace blackslash when they create their own encoding, so even nowadays many computers there would still show ¥ if you type \ into them. For instance, C:¥Users¥Username¥My Documents even when rest of the system have already supported unicode.

Don said...

Yes, that's the page on Bill's site that I was referring to. Any idea if the old FontForge utility still works? Or if there are more updated utilities for conversion of fonts (granted, the period of highest demand for such has passed).

Thanks for the SIL ref. They (as you know) had a big effort some years ago to convert materials in their own 8-bit fonts to Unicode fonts.

Don said...

Thanks c9. Sounds like the keyboard drivers in Japan/Korea have retained the assignment of the \ key to call up the ¥. In the case of Bambara Arial and the like, I believe it's a glyph change in the font (e.g., "q" replaced with "ɛ"), rather than a keyboard driver change calling up a modification of the glyph of a more obscure character (e.g., "ƒ" replaced with "ɛ"), which may have been the case with the AlphAfrica font. In any event, there was never any widespread common substitution for characters particular to many languages of West Africa such as what you describe for "\" to "¥" in Japan and Korea.