The file today

c?§U K¯K?U

Opened on any machine without the original font. Nothing is damaged and nothing is missing — these are the characters the file has always held.

After conversion

ሰላም ለዓለም

Real Unicode Amharic. It reads the same on a phone, in a browser, in an email, on a machine that has never had an Ethiopic font installed.

Both panels are the same two words — selam le'alem, “hello world”. The broken sample is illustrative: which Latin characters appear depends on which font the document was typed in.

What actually happened

Nothing was corrupted, and nothing was lost. The document is doing exactly what it was built to do — it just needs a machine from 1998 to do it on.

The old software wrote Latin letters
When a typist pressed the keys for ሰላም, the file did not receive Amharic. It received ordinary Latin letters and punctuation — whichever characters happened to sit under those keys. That is all a computer of that era could store.
The font drew Amharic on top
The legacy font had one job: wherever it saw a particular Latin character, paint an Ethiopic shape instead. On a machine with that font installed, the document looks like perfect Amharic. The letters underneath never changed.
Anywhere else, the disguise falls away
Attach it to an email, open it on a phone, upload it to a portal, hand it to a printer, or search it — and the Latin letters show through. The document is intact. What is missing is a font that has not shipped with anything for a very long time.

Decades of files, in one format nobody else can read

Ministries, universities and newspapers in Ethiopia have been typing Amharic on computers since the early 1990s. Almost all of it went into these encodings, because for most of that time there was no alternative. Archives, minutes, curricula, registries, back issues, correspondence — decades of it, held in a format that is legible on the machines it was typed on and nowhere else.

Every year that goes by, more of those machines are replaced and more of the documents stop opening properly. The text is not lost. It is one conversion away from being ordinary, searchable, forwardable Unicode.

Which encodings are read

A document does not announce which font it was typed in, so the converter works out the encoding from the file rather than asking the person holding it. These are the families it knows.

PowerGeez
the standard in most offices since the 1990s
VisualGeez
widely used in publishing and newsrooms
VG2000
the later VisualGeez generation
Nyala/NCI
the pre-Unicode NCI layout, before Nyala became a Unicode font
NCI2000
the later NCI generation
Alpas
common on older university machines
Agafari
an early Ethiopic word-processing family
Alex
an early Ethiopic word-processing family
Fedel
an early Ethiopic word-processing family
Samawerfa
an early Ethiopic word-processing family
Brana
an early Ethiopic word-processing family

14 encodings are mapped in total. 3 of them are not named here yet — they will be listed once the converter ships and the manifest is final.

How far the mapping reaches

287

Ethiopic characters the mapping tables cover, one by one

87.1%

of the core Ethiopic syllable block in Unicode

The remaining share is not a gap in the converter. Those are Ethiopic characters the legacy encodings never had a key for in the first place — so no document typed in them can contain one. Within what those encodings were able to express, the mapping is the whole of it.

The same tables serve Tigrinya as well as Amharic. Both are written in the same script, and the old fonts made no distinction between them either.

Tell me when the converter ships

The converter is not finished, so there is nothing to download here yet. Leave an address and you will hear once it is.

One message, when the converter is released. No newsletter, and the address is not passed on.