Re: Latin ligatures and Unicode

From: Michael Everson (
Date: Mon Dec 20 1999 - 07:25:46 EST

Ar 02:04 -0800 1999-12-20, scríobh Christopher John Fynn:

>IMO the best way of handling ligatures for "Fraktur" typefaces
>is by creating OpenType (or Apple's AAT) fonts with the proper
>table information. If the combinations of characters you list
>*must* be displayed by ligatures when text is set in your
>Fraktur font you can specify this in the OpenType substitution
>tables or you can make the substitution optional.

But on-or-off substitution, or even two or three level substitution, is not
sufficient to describe completely the use of ligatures in printed texts. A
font may have a repertoire of 12 ligatures, but a document may use only
two, or three, or four, or seven of them. Or all twelve. THIS is
font-specific, and can't be predicted. The only thing that could work well
is to have the font's glyph tables contain all the glyphs as X-ZWL-Y
triplets, to be displayed whenever that sequence appeared in the text.

>The substitution (whether ligatures are formed or not) may
>be specified on a language by language basis or you can
>make it the default (non-language specific) behaviour for
>your font.

This is not precise enough.

>(Of course all this will only work in OT or AAT savvy applications.)

Name one?

>I'd handle all ligatures glyphs this way even in the cases
>where a ligature character exists in Unicode. I think it best
>to make the primary way of forming ligatures through
>OpenType layout tables rather than relying on the user to
>enter a precomposed-ligature character - or the proposed
>"zero-width ligator" character.

It doesn't work well enough. I bet you anything people will start encoding
their precomposed ligatures in the Private Use Zone so they can get what
they want. This is what we ALL do NOW, with 8-bit fonts. And there is no
standard for what ligatures and what code positions, and it screws up all
operations like spelling and hyphenation, and we do it ANYWAY because it's
the only solution we have.

>A place where zero-width ligator (or non-ligator) characters
>might be required is where the presence or abscence of
>a ligature affects the meaning or other information being
>conveyed by the text.

Why? This isn't so for Arabic or Devanagari. The Virama is used
productively to produce ligation effects in Devanagari, availing of a
glyph-table triplet(+) system as described above.

>For instance applications where
>different manuscripts or editions are being compared
>there may be a real need to encode exactly where ligatures
>occur in on example of a text but not in another.

Michael Everson ** Everson Gunn Teoranta **
15 Port Chaeimhghein Íochtarach; Baile Átha Cliath 2; Éire/Ireland
Vox +353 1 478 2597 ** Fax +353 1 478 2597 ** Mob +353 86 807 9169
27 Páirc an Fhéithlinn; Baile an Bhóthair; Co. Átha Cliath; Éire

This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:20:56 EDT