The concept of mark attachment and its ideal use cases


After reading through the article on mark attachment I had some questions: It seems that this is the best way to have support for large character sets without huge file sizes, ridiculous glyph numbers, and complicated kerning groups, but in the ideal situation of full support, how would this be used?

Let’s say I’m typing in my native language, Norwegian, and I want to type the letter å. Would this involve me having to use two keystrokes instead of one? (The lowercase a followed by combining ring above). This seems especially apparent in that fonts without the legacy characters will not generate a with combining ring above when the å key is pressed.

Another question involves differentiating between letters with diacritics and ones without. The letter æ is obviously not so it is still a separate glyph in the font, but what about ø? Is it an o with a diagonal line attached or another separated glyph? I assume it gets even harder to differentiate, the more languages the font supports.

My final question is regarding which base letters to add anchors to. Is it simply the 26 base letters or do I include non diacritic letters like ß, æ, œ, etc.?

Hopefully you can enlighten me!


Do you know »unicode«? You might google a bit and find many answers to most of your questions, because unicode is the basic mechanism today to map a keystroke to a glyph representation in a text. Most of your mentioned glyphs/letters do have a unicode and hence are typeable with one keystroke (if your keyboard has it (a Norwegian keyboard will have the ø I assume, I can type it on my German layout with [opt+o])).

To your final question: all glyphs that can have certain diacritics attached to it, might want to have anchors. It doesn’t hurt to have anchors everywhere, but since for instance the ß will never have any diacritic marks, there’s no need to decorate it with any anchors.


Yes I am well aware of the unicode standard. What I meant was to ask if the ideal situation is that all regular diacritic keys would be replaced with combining ones, and letters like å would be typed with a + combining ring above (U+0061 + U+030A) as a single key instead of calling the aring letter (U+00E5) when it is pressed.


You still need all the precomposed glyphs. Software should be able to decompose them into the base glyph and combining accent but that doesn’t work reliably. And kerning doesn’t become easier without them, either. You would need a lot context kerning to deal with T + o + dieresiscomb.


I see. Thank you, Mr. Seifert!