Typing Burmese: Zawgyi, Unicode, and a decade of broken text

Text that was not what it said it was

Unicode assigns a block of code points to the Myanmar script, and each one has a defined meaning. Zawgyi used that same block and assigned its own meanings to the points instead.

The consequence is the worst one an encoding can have. Zawgyi text and Unicode text are not merely different: they are mutually unreadable while claiming to be the same thing. A document says it is Myanmar-script Unicode, software believes it, and what appears is nonsense — with no error, no warning, and nothing to indicate which of the two you are looking at except reading it.

This was not a corner case. For roughly a decade most Burmese digital text was Zawgyi, and Myanmar became the only country in the world where a majority of online text was in a non-standard encoding.

Why it happened, and what it cost

It happened for an understandable reason. Burmese is a stacking script — consonants combine, marks attach above and below, and a single syllable can be several characters deep — and rendering that correctly needs font technology that arrived later than the demand for typing Burmese did. Zawgyi sidestepped the problem by storing the shapes rather than the letters, and it worked, on the machine it was typed on.

The costs are the ones every private encoding has, and this site describes the same set for Kruti Dev in Hindi. Text cannot be searched, because the bytes are not the letters. It cannot be sorted, spellchecked, or read aloud. It cannot be safely moved between systems. And a document is only legible on a machine that has the same font, which makes every file a hostage to a font.

Myanmar declared a national migration to Unicode in October 2019. The transition is largely complete, and Zawgyi text and Zawgyi habits persist — which is why this is still worth knowing rather than a closed chapter.

There is a general lesson underneath, and it is the same one the Hindi guide draws. When a script is hard to render, somebody always ships a font that stores pictures of letters instead of letters. It solves the immediate problem and hands the language a decade of unsearchable text.

What to type now

Unicode, without hesitation. It is the standard, it is what every modern system expects, and it is what these lessons use. Myanmar3 is the common Unicode-based keyboard arrangement.

The practical thing to recognise is what a mismatch looks like. Burmese that renders as nonsense — marks in the wrong places, syllables that do not resolve, characters that look mangled rather than merely unfamiliar — is usually a Zawgyi document being shown in a Unicode font or the reverse. It is not a broken file and not a font that failed to load. Converters exist because this is common enough to need them.

If you are learning to type Burmese now, none of this affects your hands. It affects what you do with text other people send you, and what you can assume about anything written before about 2020.

The script itself

Burmese writes without spaces between words, like Thai, Lao and Khmer. Spaces separate phrases instead, so the space bar does a different job from the one it does in English and word boundaries are the reader's problem.

It stacks and combines much like Khmer: consonants take subscript forms, vowel marks attach around the letter they belong to, and what looks like one character on screen is often several in the text. This is exactly the property that made rendering hard in the first place.

It also means a keyboard chart of unshifted keys shows a part rather than the whole. The Myanmar3 arrangement uses the shifted level heavily, so the three rows the lessons draw reach most of what the practice words need and not all of it — which is the honest state of every stacking script on this site.