Microsoft limits the ‘simple’ script Latin to feature. That is ample for the Dual-script solution. You transliterate the script into Latin taking care that it goes close to the English keyboard. Next, you use orthography to make the tables. You must be well versed in the script to make the tables. I know Singhala. Devanagari is not as complex as Singhala.

I forgot to say that romanized Singhala and Unicode Singhala can be round-trip converted in either direction. If the smartfonts are hosted by Google, then there is no need for web sites to supply the downloadable font like I do. As you see, transliteration obviates the need for the limping double-byte solution. (You can read Singhala as Latin script or native script, convenient for my daughter who was born in US and never learned Sinhala writing. The military and spy agencies may get other ideas too, such as voice to text).

Thanks.

JC

These comments apply to Indic, but also sort of apply to Arabic too. Indic user group is by no measure a ‘small subset of the population’. One fifth of humanity speak Indic. As you say, the scripts are ‘lesser-known scripts’ because of their bad implementation on the computer. Unicode double-byte that adopted Indian ISCII trumped singe-byte ISO-8859 idea. More reasons later…

I developed a set of test smartfonts for Singhala in 2004. At that time, only SIL’s WorldPad was able to display it. They have only lookup tables.
Here are the two web sites that use one of these fonts:
http://www.lovatasinhala.com/ (hand coded)
http://www.ahangama.com/ (WordPress blog)
The font actually implements romanized Singhala in the native script by means of the feature. It is intuitive to type. It needs no double-byte code supporting software, only OpenType aware apps. If NBH is implemented, it could be typed mixing two orthographies never needing any double-byte characters (specifically, ZWNJ). Before jumping into conclusions, please recall that the underlying code set used for Fraktur is Latin, though the letter shapes are different.

The smartfont debunks the notion that Indic is so complex that it would always be a limping script on the computer (Abugida!). The approach taken by Unicode to digitize Indic made it very awkward on the computer. Counting Indic web sites, we see how terrible it must be.

Unicode applied the same principle to Indic as used to define the Latin script. That is, a script is a set of letter symbols for writing a script, with no significance in the grammar of the language. That is well and good for languages using the Latin script because letters hardly change shape in combination. (English has only may be five optional ligatures and Fraktur has a few more)

Unicode solution for Indic is complex. The fonts should make internal shapes that are not necessarily letters of the script but component parts of complex letters, and then write internal routines that implement their assembly. Typing Unicode Singhala, one of the more complex of Indic scripts, is very frustrating. It is forcing us forget our orthographies and is interfered by habits with the English keyboard. This is unfortunate because the computer overcomes all difficulties of letterpress printing and yet we cannot implement Indic with that happy outcome.

As a native user of Singhala, I know that we write by first thinking of the sound and then putting down its shape. In contrast, with English, we write the dictionary spelling of a word.

In Sanskrit grammar, the first and the foundation lesson is the description of ‘akshara’. The Sanskrit akshara set (hodiya) is extended to form the hodiya of each Indic language. The sound (shabda) of an akshara defines the letter as we conceive it. It is the phoneme or as we call, the vowel or consonant. The shape of the akshara called ‘ruupa’ depends on the local script (Nagari, Tamil or Singhala?). The script has its specific convention(s) of writing or orthography. A perfect application of this principle is the Harvard-Kyoto Sanskrit alphabet:
http://www.sanskrit-lexicon.uni-koeln.de/scans/MWScan/tamil/index.html
It is Sanskrit implemented in the Latin script and has no orthographic constraints.
Notice that the chart (hodiya) is in table form unlike an alphabet. It is like IPA chart in its original form.

The problem was to find a way to show Singhala text that closely matches its orthography. Singhala script has three orthographies — one each for Singhala, Sanskrit and Pali (correctly, Magadhi). Modern Singhala mixes Singhala and Sanskrit. Pali is written separately. When you write Singhala, that always mixes in Sanskrit, the Sanskrit words combine letters according to the rule set of Sanskrit. We should either make separate smartfonts for each orthography or use the more complex font and use ZWNJ to prevent conjoints inside the less complex Sinhala words. (few and far apart).

This solution could also be called the dual-script solution for Indic. That makes OCR simple making it possible to digitize old Indic books.

JC
ahangama@gmail.com
817-473-9984

Also, congratulations on shipping Firefox 10.

https://bugzilla.mozilla.org/