The name table: where every name comes from

The name shown in an interface does not come from the file name, and it is not guessed by the system. It comes from the name table inside the font: a list of "number → string" records, where the number says what kind of name the string is.

What the table contains#

The layout is plain:

• a header with a version and a record count;
• one record per string: platform, encoding, language, name ID, length, offset;
• the strings themselves, stored after the records.

Which means the same name can appear many times: one record in Chinese, one in English, one for the Macintosh platform, one for Windows. Which one wins depends on the program reading it.

The name IDs that matter#

The difference between 1 + 2 and 16 + 17 is the part people skip. Old software only knows 1 and 2, so a "Inter Bold" face may store family Inter Bold and subfamily Regular; 16 and 17 express the design intent instead — "this belongs to Inter, and it is the Bold member". The first pair is compatibility, the second is truth.

Platforms and languages#

Every record carries a platform / encoding / language triple. Two groups cover almost everything:

• Windows (platformID 3): encoding 1 = BMP, 10 = full Unicode; language 0x0409 English, 0x0804 Simplified Chinese, 0x0404 Traditional Chinese;
• Macintosh (platformID 1): language 0 for English, other Mac language codes otherwise.

Older fonts sometimes wrote names only for the Macintosh platform, leaving the Windows group with nothing but a copyright or licence string. Picking one platform and reading only that will lose the name; fall back field by field instead.

Localised names#

A Chinese user seeing Microsoft YaHei pauses for a second. The table usually contains a Chinese record already (a Chinese entry for nameID 1 or 16); many programs simply take the English one.

This site stores the localised entries of nameID 1 / 16 as localizedNames, shows the local name in family lists, and keeps the real name in the tooltip. Search, selection and favourites always use the real name — otherwise "雅黑" and "YaHei" would produce two different result sets.

Three pitfalls#

1. The PostScript name is tightly restricted: at most 63 characters, ASCII 33–126, no spaces and none of [](){}<>/%. It ends up in PDFs and system font caches; a bad value is accepted by some systems and quietly breaks others.
2. WWS is not the typographic family. WWS (weight, width, slant) exists so software can group faces mechanically; the typographic names preserve the designer's grouping. When the two disagree, different systems show different names.
3. Names can lie. The table is just text with no validation; a file can claim its family is Arial. That is why grouping also looks at other fields.

How this site uses it#

• Family list and detail heading: the family name from name (typographic family first, plain family as fallback);
• Localised display when the interface language changes: the other language records of the same nameID;
• Search aliases: localised names and common Chinese names flattened into one alias table;
• Numbers such as glyph count are not here — they come from maxp and OS/2.

Further reading#

• OpenType spec: the name table, including the full nameID list and platform encodings
• OpenType spec: the naming section covering WWS versus typographic names