Dynamically generating Ge'ez unicodes

Your question asks "how to create a language", so I will describe all the pieces that need to be in place for a new language (or more accurately, writing system). You ask specifically about the Eritrean alphabet, so I will provide specific examples of how that is supported on modern systems, and try to provide you pointers for the pieces you are missing. The answer is long, and provides lots of links, to support the two explanations.

To work with a script like Ge'ez (also known as Ethiopic, the script used to write Amharic in Ethiopia and Tigrinya in Eritrea) you need a few things. The first is a way to encode the characters; a set of numbers representing each character, that the computer can use to represent the text. Luckily, Unicode has become widespread, and Unicode is designed to be a universal character set that includes all of the world's languages. Unicode 3.0 introduced Ethiopic in the range U+1200-U+137F, and later versions added supplements of more obscure characters in the ranges U+1380-U+1394, U+2D80-U+2DDF and U+AB00-U+AB2F. If you wanted to support a language that Unicode didn't yet support, you would either need to use the private use area and define your own mapping of characters to code points, or submit a proposal to have your script added to Unicode; for example, see the proposal for Ethiopic.

Now, Unicode is just a character set; an abstract mapping between characters and numbers. To actually transmit these characters as a sequence of bytes, you use a character encoding. There are many encodings; some of them, like ASCII and ISO-8859-1 only cover a subset of the full Unicode character set, while others, like UTF-8 and UTF-16, cover the full range. For documents on the web, UTF-8 is the recommended character encoding; you should never use anything else if you can help it. In UTF-8, you can write Ge'ez directly in the document, for example: ኤርትራ. One thing to watch out for is that some programs (especially on Windows) will offer you "Unicode" as an encoding, when they mean UTF-16; you want to make sure to choose UTF-8, as it's more efficient and more compatible with a wider variety of software.

If you are using encodings that don't cover the full range of Unicode, or you don't have a good way to type those characters, and you are writing HTML or XML, you can use numeric character references instead. To do this, you write the Unicode code point of the character you want to refer between &# and ;. You can write the number in decimal, or in hexadecimal prefixed with an x. For example, ሀ can be written ሀ or ሀ (the semicolon at the end is important; it wasn't working for you in the comments because you were missing it).

Now that you have a character set, and a way of encoding it, you need a way to display it. Some scripts are easier to display in others. For all scripts, you need a font; a file defining how each character looks. A font contains a collection of glyphs, or drawings of each character. Some scripts, like the Latin alphabet (the alphabet used for English and most European languages) are relatively simple; each character is a separate glyph, and how they are drawn doesn't depend on what characters come before or after (though diacritics and ligatures can make it a little more complicated). Others, like Arabic and Indic scripts are written in cursive, where letters join to each other so how they are drawn can depend on the characters near them. These languages require special rendering support like Uniscribe or DirectWrite on Windows, Pango on Linux, or advanced font technology like Apple Advanced Typography or Graphite.

Luckily, Ge'ez is a fairly simple writing system, that doesn't require any specialized rending support or advanced font systems. Each of the characters is a separate glyph, and it doesn't require any reordering. So a normal OpenType font, displayed with the rendering systems already available on most computers, will do the job. But you still need the font in order to be able to display the characters. To create you own font, you can use FontForge (a free/open source tool), Fontographer, FontLab Studio, or other similar software.

For Ethiopic, you don't need to create your own. There are numerous fonts available that include the Ethiopic characters, but one that I would recommend is Abyssinica SIL from SIL (the Summer Institute of Linguistics), which does a lot of great work for minority languages and writing systems. Their fonts are available under a free license, that allows you to use the font, redistribute the font, and modify the font, so their fonts are quite flexible and can be used in a wide variety of situations. Windows ships with Nyala, which includes Ethiopic characters, since Windows Vista, and Ebrima, which added support for Ethiopic characters in Windows 8; so people on Windows Vista or later should be able to view Ethiopic characters already. Mac OS X ships with Kefa as of 10.6.

Once you have the font, you will be able to view Ethiopic characters. But other people reading your documents might not have those fonts (if they are using an older version of Windows or Mac OS X, if they didn't install all of the fonts that came with Windows, or the like), in which case the characters will probably show up as boxes or question marks on their machine. You could give those people a redistributable font like Abyssinica SIL, or they could buy a font that includes Ethiopic characters, but that can be inconvenient. For working with word processor documents or plain text, that's probably the best you can do; they will need the font installed on their computer to be able to display the text. If you create a PDF on your computer, it should embed the fonts that it needs to display the text, so creating a PDF can be a convenient way to include uncommon fonts with your document.

On a web page, you can use web fonts to link to a font from your stylesheet, allowing the users web browser to load that font for that web page. Web fonts are supported all the way back to IE 6, and in recent versions of most other web browsers, so they are actually quite widely supported. Different web browsers support different font file formats (EOT, TTF, OpenType, SVG, and WOFF), and slightly different syntaxes for the CSS (older versions of IE are based on an older draft), so it can be a bit tricky to make a page that is compatible with all browsers. Luckily, people have automated that process. Some web fonts are available online from Google Web Fonts or FontSquirrel, but sadly, I couldn't find any Ethiopic fonts already hosted. However, you can upload a font to FontSquirrel, and it will convert it into all of the major formats, and provide example CSS that will work on all modern browsers. Note that you should only do this with fonts that allow web embedding; not all fonts do. Since Abyssinica SIL is available under the Open Font License, you can use it, and I've run it through FontSquirrel for you; you can see how it works (check out the Glyphs & Languages tab), or download the kit. To use it, just put the font files (.ttf, .eot, .svg, and .woff) on your server in the same directory as your CSS, and include the following in your CSS:

@font-face {
    font-family: 'abyssinica_silregular';
    src: url('abyssinicasil-r.eot');
    src: url('abyssinicasil-r.eot?#iefix') format('embedded-opentype'),
         url('abyssinicasil-r.woff') format('woff'),
         url('abyssinicasil-r.ttf') format('truetype'),
         url('abyssinicasil-r.svg#abyssinica_silregular') format('svg');
    font-weight: normal;
    font-style: normal;
}

Now that you know how to encode Ethiopic, view Ethiopic characters, and share documents containing Ethiopic characters, you are probably going to want to type them into documents. If you are using HTML, you could just type the numeric character reference described above. In other documents, you could just copy and paste the characters from a chart of all of them, like the Wikipedia page. But that would become pretty cumbersome. Depending on your system and settings, you can also use Unicode Hex Input to enter arbitrary Unicode characters, but that is also cumbersome.

To fully support typing a script on your computer, you need a keyboard layout or input method. Some scripts can be typed with a simple keyboard layout, which says which keys correspond to which characters. If a script has more characters than there are keys on the keyboard, Shift and Alt (or Option on the Mac) can be used to map to more characters. Dead keys can also be used to expand the range of characters that you type; dead keys are sequences of two or more keystrokes that produce a single glyph; for example, on Mac OS X, to type "á", you can type Option-E A. To create a keyboard layout on Windows, you can use the Microsoft Keyboard Layout Creator. Mac OS X uses an XML format for keyboard layouts, so you can create one directly, or use Ukelele from SIL to create one more easily. On systems using X11 (like Linux), you can create your own XKB layouts.

If you need more characters than can be supported with modifiers and dead keys, like typing Chinese or Japanese, then you need a full-fledged input method. An input method allows you to run arbitrary code to map what someone types into the text it produces; for example, in a Japanese input method, you may type a phonetic representation of what you you are writing, and it will show you a drop down list of possible characters that match that representation, allowing you to choose the appropriate ones. Windows provides the Input Method Manager for writing input methods, Mac OS X the Input Method Kit, and X11 has a few ways to do it, such as SCIM and iBus.

The standard input method for Ethiopic makes extensive use of dead keys. It looks like the most popular existing input method for Ethiopic is Keyman, which is a commercial input method that works on Mac and Windows, and in addition there's a free variant, KMFL, that works on Linux. SIL has keyboard downloads for this input method; they also have a keyboard layout for Mac OS X which uses dead keys to achieve the same thing. Mac OS X has more extensive dead key support, so it doesn't require an input method to support this form of input, while on Windows you need to use an input method like Keyman to be able to enter input this way. Google has a free input method for Windows, Google Input Tools for Windows, which supports Amharic, and allows you to customize its input schemes; you could try adapting their Amharic support for Tigrinya.

If you just need to support input on a web site, you could do this in JavaScript, by writing an input method in JavaScript that transliterates from what someone types into Ethiopic. I do not know of any existing frameworks for doing this; however, I have found Korean and Japanese input methods implemented in JavaScript. You could take a look at how those are implemented. Upon looking further, I've found that Tavultesoft, who make Keyman, also have KeymanWeb, a JavaScript based input method that you can buy and embed in your site. MediaWiki also has an input method extension Narayam, that includes a JavaScript based input method for MediaWiki based sites like Wikipedia, which includes an experimental Amharic input method. There is also a draft W3C IME API, which helps provide an interface between web apps and native IMEs, as well as JavaScript based IMEs. Given that it's still a draft, I don't know if it is yet supported anywhere.

With all the above (a character set, encoding, fonts, rendering support, and an input method), you will be able to create, share, and view documents in your script. If that's all you need, great; the above will allow you to work with documents in a given script. But for full support for a language on your computer, not just its script or writing system, there are two more pieces that you need: a locale, and your software to be localized (translated and adapted) for your language.

A locale specifies how programs should manipulate text in a given script, language, culture, and/or encoding. There are many common text processing operations that programs do: displaying numbers, displaying dates and times, sorting strings or names, and so on. How these should work can differ based on the language, script, and culture of the person using the program; for instance, in Swedish "ü" is sorted along with "y", while in English and German it's sorted along with "u". Differences may not be based on language: both Mexico and Spain use Spanish, but in Mexico numbers are displayed with . as the decimal separator (1½ is written "1.5"), while in Spain , is used as the decimal separator (1½ is written "1,5"). A locale specifies all of these rules. Because the locale can vary based on language, culture, and sometimes other factors, the language and country are usually used to specify the locale, and other information can be used as well.

The most widely used standard for naming locales is RFC 4646 (BCP 47). Locales are usually specified as "ln-CC" with the language code ln and country code CC: US English is en-US, British English is en-UK, and French in France is fr-FR. If more information needs to be specified, it can be included. For instance, Serbian can be written with either Latin or Cyrillic, and so Serbian in Serbia can be either sr-Latn-CS or sr-Cyrl-CS. Tigrinya in Eritrea is written ti-ER.

There are a variety of different formats for defining the rules that a particular locale has. Windows uses NLP files, a custom format that can be created with Microsoft Locale Builder. POSIX (Unix/Linux) locales can be created using localedef. Many systems these days are moving towards the Unicode Common Locale Data Registry, which specifies a standardized format for locale data as well as a comprehensive database of locales for many of the worlds languages. ICU is a library for C and Java (and used by many other environments) for manipulating Unicode text according to Unicode rules and locale data; they have a good browser for the data from the CLDR and their own locale data. For example, take a look at their entry for ti-ER.

Finally, for full support of a language, you need to translate the software itself into that language. There are, of course, many pieces of software, and each one contains many strings that need to be translated. Some software is not designed to be translated; it has not been internationalized. Some software can only be translated by whoever created it; the strings are built into the program and cannot be easily modified by a third party. But it is possible to localize some software, translating it to your language and culture. If the software has already been localized for several other languages and cultures, it is likely to be flexible enough to support a new language, and if it uses formats that are easily modifiable for localization information, it can be modified by third parties.

For instance, applications on Mac OS X store their localization data in separate files within the application bundle. There is a tool called AppleGlot (you need to register for the Mac Developer Program and go to the downloads area to find it) which can help you extract that data, provide a file with all of the strings which need to be translated, and allow you to combine that with the application again once you have. For open source software, such as much software available on Linux, you can work with the developers to provide translation. Some software uses gettext for translation strings, which use the PO file format that you can edit using poedit. Some uses Qt, for which you can use Qt Linguist. Or for dealing with a wide variety of formats, you can use a commercial offering like Swordfish or Transifex.

Of course, no one person can do all of the above; it takes many people working together to build support for a new language on modern computer systems. This is all intended to be a high-level tour of all of the components that go into language support for a given language, with references that will help you follow up on whichever aspects you would like to work on, as well as demonstrate what already works for Tigrinya and the Ge'ez script.

If they are Unicode characters they should be displayable just like characters of any other language. I googled it and found this, hopefully they're the same ones you're asking about:

የ ዩ ዪ ያ ዬ ይ ዮ

ዸ ዺ ዻ ዼ ዽ ዾ

See? No extra work required to display them on web browsers or other programs.

---

These are characters from the Unicode Ethiopic set (U+1200..U+137C), encoded in UTF-8:

Line 1:

• የ = 0xE1 0x8B 0xA8 = U+12E8 = ETHIOPIC SYLLABLE YA
• ዩ = 0xE1 0x8B 0xA9 = U+12E9 = ETHIOPIC SYLLABLE YU
• ዪ = 0xE1 0x8B 0xAA = U+12EA = ETHIOPIC SYLLABLE YI
• ያ = 0xE1 0x8B 0xAB = U+12EB = ETHIOPIC SYLLABLE YAA
• ዬ = 0xE1 0x8B 0xAC = U+12EC = ETHIOPIC SYLLABLE YEE
• ይ = 0xE1 0x8B 0xAD = U+12ED = ETHIOPIC SYLLABLE YE
• ዮ = 0xE1 0x8B 0xAE = U+12EE = ETHIOPIC SYLLABLE YO

Line 2:

• ዸ = 0xE1 0x8B 0xB8 = U+12F8 = ETHIOPIC SYLLABLE DDA
• ዺ = 0xE1 0x8B 0xBA = U+12FA = ETHIOPIC SYLLABLE DDI
• ዻ = 0xE1 0x8B 0xBB = U+12FB = ETHIOPIC SYLLABLE DDAA
• ዼ = 0xE1 0x8B 0xBC = U+12FC = ETHIOPIC SYLLABLE DDEE
• ዽ = 0xE1 0x8B 0xBD = U+12FD = ETHIOPIC SYLLABLE DDE
• ዾ = 0xE1 0x8B 0xBE = U+12FE = ETHIOPIC SYLLABLE DDO

Using Ethiopian characters on web pages is mostly a matter of fonts these days. (You may also have a problem with entering them conveniently, but this depends on your authoring environmentPeople using e.g. Windows 7 have at least one font containing them, but old computers typically lack such fonts. The following fonts contain them (there may be others):

Code 2000, was freeware, the author has disappeared, so the status is obscure

Unifont, a free bitmap font

FreeSerif, a free font

Nyala, distributed with some versions of Windows

SunExt-A, a free font

Fixedsys Excelsior, a free bitmap font I suppose (haven’t tested)

I would probably use FreeSerif as a downloadable font, with @font-face.