Fonts in X11R6.7 : More about core fonts
Previous: Fonts included with X11R6.7
Next: Appendix: background and terminology

4. More about core fonts

This section describes X11R6.7-specific enhancements to the core X11 fonts system.

4.1. Core fonts and internationalisation

The scalable font backends (Type 1, Speedo and TrueType) can automatically re-encode fonts to the encoding specified in the XLFD in `fonts.dir'. For example, a `fonts.dir' file can contain entries for the Type 1 Courier font such as

cour.pfa -adobe-courier-medium-r-normal--0-0-0-0-m-0-iso8859-1 
cour.pfa -adobe-courier-medium-r-normal--0-0-0-0-m-0-iso8859-2
which will lead to the font being recoded to ISO 8859-1 and ISO 8859-2 respectively.

4.1.1. The fontenc layer

Three of the scalable backends (Type 1, Speedo, and the FreeType TrueType backend) use a common fontenc layer for font re-encoding. This allows these backends to share their encoding data, and allows simple configuration of new locales independently of font type.

Please note: the X-TrueType (X-TT) backend does not use the fontenc layer, but instead uses its own method for font reencoding. If you are only interested in X-TT you may want to skip to Using Symbol Fonts, as the intervening information does not apply to X-TT. X-TT itself is described in more detail in X-TrueType.

In the fontenc layer, an encoding is defined by a name (such as iso8859-1), possibly a number of aliases (alternate names), and an ordered collection of mappings. A mapping defines the way the encoding can be mapped into one of the target encodings known to fontenc; currently, these consist of Unicode, Adobe glyph names, and arbitrary TrueType ``cmap''s.

A number of encodings are hardwired into fontenc, and are therefore always available; the hardcoded encodings cannot easily be redefined. These include:

Additional encodings can be added by defining encoding files. When a font encoding is requested that the fontenc layer doesn't know about, the backend checks the directory in which the font file resides (not necessarily the directory with fonts.dir!) for a file named `encodings.dir'. If found, this file is scanned for the requested encoding, and the relevant encoding definition file is read in. The `mkfontdir' utility, when invoked with the `-e' option followed by the name of a directory containing encoding files, can be used to automatically build `encodings.dir' files. Please see the mkfontdir(1) manual page for more details.

A number of encoding files for common encodings are included with X11R6.7. Information on writing new encoding files can be found in Format of encodings directory files and Format of encoding files later in this document.

4.1.2. Backend-specific notes about fontenc

4.1.2.1. The FreeType backend

For TrueType and OpenType fonts, the FreeType backend scans the mappings in order. Mappings with a target of PostScript are ignored; mappings with a TrueType or Unicode target are checked against all the cmaps in the file. The first applicable mapping is used.

For Type 1 fonts, the FreeType backend first searches for a mapping with a target of PostScript. If one is found, it is used. Otherwise, the backend searches for a mapping with target Unicode, which is then composed with a built-in table mapping codes to glyph names. Note that this table only covers part of the Unicode code points that have been assigned names by Adobe.

Specifying an encoding value of adobe-fontspecific for a Type 1 font disables the encoding mechanism. This is useful with symbol and incorrectly encoded fonts (see Incorrectly encoded fonts below).

If a suitable mapping is not found, the FreeType backend defaults to ISO 8859-1.

4.1.2.2. Type 1

The Type 1 backend behaves similarly to the FreeType backend with Type 1 fonts, except that it limits all encodings to 8-bit codes.

4.1.2.3. Speedo

The Speedo backend searches for a mapping with a target of Unicode, and uses it if found. If none is found, the backend defaults to ISO 8859-1.

The Speedo backend limits all encodings to 8-bit codes.

4.1.3. Format of encoding directory files

In order to use a font in an encoding that the font backend does not know about, you need to have an `encodings.dir' file either in the same directory as the font file used or in a system-wide location (`/usr/X11R6/lib/X11/fonts/encodings/' by default).

The `encodings.dir' file has a similar format to `fonts.dir'. Its first line specifies the number of encodings, while every successive line has two columns, the name of the encoding, and the name of the encoding file; this can be relative to the current directory, or absolute. Every encoding name should agree with the encoding name defined in the encoding file. For example,

3
mulearabic-0 /usr/X11R6/lib/X11/fonts/encodings/mulearabic-0.enc
mulearabic-1 /usr/X11R6/lib/X11/fonts/encodings/mulearabic-1.enc
mulearabic-2 /usr/X11R6/lib/X11/fonts/encodings/mulearabic-2.enc

The name of an encoding must be specified in the encoding file's `STARTENCODING' or `ALIAS' line. It is not enough to create an `encodings.dir' entry.

If your platform supports it (it probably does), encoding files may be compressed or gzipped.

The `encoding.dir' files are best maintained by the `mkfontdir' utility. Please see the mkfontdir(1) manual page for more information.

4.1.4. Format of encoding files

The encoding files are ``free form,'' i.e. any string of whitespace is equivalent to a single space. Keywords are parsed in a non-case-sensitive manner, meaning that `size', `SIZE', and `SiZE' all parse as the same keyword; on the other hand, case is significant in glyph names.

Numbers can be written in decimal, as in `256', in hexadecimal, as in `0x100', or in octal, as in `0400'.

Comments are introduced by a hash sign `#'. A `#' may appear at any point in a line, and all characters following the `#' are ignored, up to the end of the line.

The encoding file starts with the definition of the name of the encoding, and possibly its alternate names (aliases):

STARTENCODING mulearabic-0
ALIAS arabic-0
The name of the encoding and its aliases should be suitable for use in an XLFD font name, and therefore contain exactly one dash `-'.

The encoding file may then optionally declare the size of the encoding. For a linear encoding (such as ISO 8859-1), the SIZE line specifies the maximum code plus one:

SIZE 0x2B
For a matrix encoding, it should specify two numbers. The first is the number of the last row plus one, the other, the highest column number plus one. In the case of `jisx0208.1990-0' (JIS X 0208(1990), double-byte encoding, high bit clear), it should be
SIZE 0x75 0x80
In the case of a matrix encoding, a `FIRSTINDEX' line may be included to specify the minimum glyph index in an encoding. The keyword `FIRSTINDEX' is followed by two integers, the minimum row number followed by the minimum column number:
FIRSTINDEX 0x20 0x20
In the case of a linear encoding, a `FIRSTINDEX' line is not very useful. If for some reason however you chose to include on, it should be followed by a single integer.

Note that in most font backends inclusion of a `FIRSTINDEX' line has the side effect of disabling default glyph generation, and this keyword should therefore be avoided unless absolutely necessary.

Codes outside the region defined by the `SIZE' and `FIRSTINDEX' lines are understood to be undefined. Encodings default to linear encoding with a size of 256 (0x100). This means that you must declare the size of all 16 bit encodings.

What follows is one or more mapping sections. A mapping section starts with a `STARTMAPPING' line stating the target of the mapping. The target may be one of:

Every line in a mapping section maps one from the encoding being defined to the target of the mapping. In mappings with a Unicode or TrueType mapping, codes are mapped to codes:
0x21 0x0660
0x22 0x0661
...
As an abbreviation, it is possible to map a contiguous range of codes in a single line. A line consisting of three integers
<it/start/ <it/end/ <it/target/
is an abbreviation for the range of lines
start target
start+1 target+1
...
end target+end-start
For example, the line
0x2121 0x215F 0x8140
is an abbreviation for
0x2121 0x8140
0x2122 0x8141
...
0x215F 0x817E
Codes not listed are assumed to map through the identity (i.e. to the same numerical value). In order to override this default mapping, you may specify a range of codes to be undefined by using an `UNDEFINE' line:
UNDEFINE 0x00 0x2A
or, for a single code,
UNDEFINE 0x1234

PostScript mappings are different. Every line in a PostScript mapping maps a code to a glyph name

0x41 A
0x42 B
...
and codes not explicitly listed are undefined.

A mapping section ends with an ENDMAPPING line

ENDMAPPING
After all the mappings have been defined, the file ends with an ENDENCODING line
ENDENCODING

In order to make future extensions to the format possible, lines starting with an unknown keyword are silently ignored, as are mapping sections with an unknown target.

4.1.5. Using symbol fonts

Type 1 symbol fonts should be installed using the adobe-fontspecific encoding.

In an ideal world, all TrueType symbol fonts would be installed using one of the microsoft-symbol and apple-roman encodings. A number of symbol fonts, however, are not marked as such; such fonts should be installed using microsoft-cp1252, or, for older fonts, microsoft-win3.1.

In order to guarantee consistent results (especially between Type 1 and TrueType versions of the same font), it is possible to define a special encoding for a given font. This has already been done for the ZapfDingbats font; see the file `encodings/adobe-dingbats.enc'.

4.1.6. Hints about using badly encoded fonts

A number of text fonts are incorrectly encoded. Incorrect encoding is sometimes done by design, in order to make a font for an exotic script appear like an ordinary Western text font on systems which are not easily extended with new locale data. It is often the result of the font designer's laziness or incompetence; for some reason, most people seem to find it easier to invent idiosyncratic glyph names rather than follow the Adobe glyph list.

There are two ways of dealing with such fonts: using them with the encoding they were designed for, and creating an ad hoc encoding file.

4.1.6.1. Using fonts with the designer's encoding

In the case of Type 1 fonts, the font designer can specify a default encoding; this encoding is requested by using the `adobe-fontspecific' encoding in the XLFD name. Sometimes, the font designer omitted to specify a reasonable default encoding, in which case you should experiment with `adobe-standard', `iso8859-1', `microsoft-cp1252', and `microsoft-win3.1'. (The encoding `microsoft-symbol' doesn't make sense for Type 1 fonts).

TrueType fonts do not have a default encoding. However, most TrueType fonts are designed with either Microsoft or Apple platforms in mind, so one of `microsoft-symbol', `microsoft-cp1252', `microsoft-win3.1', or `apple-roman' should yield reasonable results.

4.1.6.2. Specifying an ad hoc encoding file

It is always possible to define an encoding file to put the glyphs in a font in any desired order. Again, see the `encodings/adobe-dingbats.enc' file to see how this is done.

4.1.6.3. Specifying font aliases

By following the directions above, you will find yourself with a number of fonts with unusual names --- with encodings such as `adobe-fontspecific', `microsoft-win3.1' etc. In order to use these fonts with standard applications, it may be useful to remap them to their proper names.

This is done by writing a `fonts.alias' file. The format of this file is very simple: it consists of a series of lines each mapping an alias name to a font name. A `fonts.alias' file might look as follows:

"-ogonki-alamakota-medium-r-normal--0-0-0-0-p-0-iso8859-2" \
  "-ogonki-alamakota-medium-r-normal--0-0-0-0-p-0-adobe-fontspecific"
(both XLFD names on a single line). The syntax of the `fonts.alias' file is more precisely described in the mkfontdir(1) manual page.

4.2. Additional notes about scalable core fonts

The FreeType (libfreetype-xtt2) backend (module `freetype', formerly known as xfsft) is able to deal with both TrueType and Type 1 fonts. This puts it in conflict with the X-TT and Type 1 backends respectively.

If both the FreeType and the Type 1 backends are loaded, the FreeType backend will be used for Type 1 fonts. If both the FreeType and X-TT backends are loaded, X-TT will be used for TrueType fonts.

4.2.1. About the FreeType backend

The FreeType (libfreetype-xtt2) backend (formerly xfsft) is a backend based on version 2 of the FreeType library (see the FreeType web site) and has the X-TT functionalities for CJKV support provided by the After X-TT Project (see the After X-TT Project web site). The FreeType module has support for the ``fontenc'' style of internationalisation (see The fontenc layer). This backend supports TrueType font files (`*.ttf'), OpenType font files (`*.otf'), TrueType Collections (`*.ttc'), OpenType Collections (`*.otc') and Type 1 font files (`*.pfa' and `*.pfb').

In order to access the faces in a TrueType Collection file, the face number must be specified in the fonts.dir file before the filename, within a pair of colons, or by setting the 'fn' TTCap option. For example,

:1:mincho.ttc -misc-pmincho-medium-r-normal--0-0-0-0-p-0-jisx0208.1990-0

refers to face 1 in the `mincho.ttc' TrueType Collection file.

The new FreeType backend supports the extended `fonts.dir' syntax introduced by X-TrueType with a number of options, collectively known as `TTCap'. A `TTCap' entry follows the general syntax

option=value:

and should be specified before the filename. The new FreeType almost perfectly supports TTCap options that are compatible with X-TT 1.4. The Automatic Italic (`ai'), Double Strike (`ds') and Bounding box Width (`bw') options are indispensable in CJKV. For example,

mincho.ttc -misc-mincho-medium-r-normal--0-0-0-0-c-0-jisx0208.1990-0
ds=y:mincho.ttc -misc-mincho-bold-r-normal--0-0-0-0-c-0-jisx0208.1990-0
ai=0.2:mincho.ttc -misc-mincho-medium-i-normal--0-0-0-0-c-0-jisx0208.1990-0
ds=y:ai=0.2:mincho.ttc -misc-mincho-bold-i-normal--0-0-0-0-c-0-jisx0208.1990-0
bw=0.5:mincho.ttc -misc-mincho-medium-r-normal--0-0-0-0-c-0-jisx0201.1976-0
bw=0.5:ds=y:mincho.ttc -misc-mincho-bold-r-normal--0-0-0-0-c-0-jisx0201.1976-0
bw=0.5:ai=0.2:mincho.ttc -misc-mincho-medium-i-normal--0-0-0-0-c-0-jisx0201.1976-0
bw=0.5:ds=y:ai=0.2:mincho.ttc -misc-mincho-bold-i-normal--0-0-0-0-c-0-jisx0201.1976-0

setup the complete combination of jisx0208 and jisx0201 using mincho.ttc only. More information on the TTCap syntax is found on the After X-TT Project page.

The FreeType backend uses the fontenc layer in order to support recoding of fonts; this was described in The fontenc layer and especially FreeType-specific notes about fontenc earlier in this document.

4.2.2. About the X-TrueType TrueType backend

The `X-TrueType' backend is a backend based on version 1 of the FreeType library. X-TrueType doesn't use the `fontenc' layer for managing font encodings, but instead uses its own database of encodings.

Since the functionalities for CJKV support introduced by X-TT have been merged into the new FreeType backend, the X-TT backend will be removed from X11R6.7's tree near the future. Therefore, the use of FreeType backend is preferred over the X-TT backend.

General information on X-TrueType may be found at the After X-TT Project page.

4.2.3. Delayed glyph rasterisation

When loading a proportional fonts which contain a huge number of glyphs, the old FreeType delayed glyph rasterisation until the time at which the glyph was first used. The new FreeType (libfreetype-xtt2) has an improved `very lazy' metric calculation method to speed up the process when loading TrueType or OpenType fonts. Although the X-TT module also has this method, the "vl=y" TTCap option must be set if you want to use it. This is the default method for FreeType when it loads multi-byte fonts. Even if you use a unicode font which has tens of thousands of glyphs, this delay will not be worrisome as long as you use the new FreeType backend -- its `very lazy' method is super-fast.

The maximum error of bitmap position using `very lazy' method is 1 pixel, and is the same as that of a character-cell spacing. When the X-TT backend is used with the `vl=y' option, a chipped bitmap is displayed with certain fonts. However, the new FreeType backend has minimal problem with this, since it corrects left- and right-side bearings using `italicAngle' in the TrueType/OpenType post table, and does automatic correction of bitmap positions when rasterisation so that chipped bitmaps are not displayed. Nevertheless if you don't want to use the `very lazy' method when using multi-bytes fonts, set `vl=n' in the TTCap option to disable it:

vl=n:luxirr.ttf -b&h-Luxi Serif-medium-r-normal--0-0-0-0-p-0-iso10646-1

Of course, both backends also support an optimisation for character-cell fonts (fonts with all glyph metrics equal, or terminal fonts). A font with an XLFD specifying a character-cell spacing `c', as in

-misc-mincho-medium-r-normal--0-0-0-0-c-0-jisx0208.1990-0

or

fs=c:mincho.ttc -misc-mincho-medium-r-normal--0-0-0-0-p-0-jisx0208.1990-0

will not compute the metric for each glyph, but instead trust the font to be a character-cell font. You are encouraged to make use of this optimisation when useful, but be warned that not all monospaced fonts are character-cell fonts.


Fonts in X11R6.7 : More about core fonts
Previous: Fonts included with X11R6.7
Next: Appendix: background and terminology