Differences between revisions 94 and 95
Revision 94 as of 2006-03-16 08:20:37
Size: 23199
Editor: EddyPetrisor
Comment: Update the CJK font status
Revision 95 as of 2006-03-16 08:24:38
Size: 23259
Editor: EddyPetrisor
Comment: Minor changes
Deletions are marked like this. Additions are marked like this.
Line 200: Line 200:
It seems that for arabic, nazli is the most fitted for arabic scripts.

Line 224: Line 227:
||<rowbgcolor="#ffffaa"> ja ||Japanese||ttf-cjk-compact-udeb||ttf-sazanami-gothic||Kenshi and Hidetaka recommend Sazanami-Gothic. A stripped ttf is about 160KB. Still there is a [http://lists.debian.org/debian-boot/2006/02/msg00936.html question] if the kana and kanji glyphs are well mixed for Japanese.|| ||<rowbgcolor="#ffffaa"> ja ||Japanese||ttf-cjk-compact-udeb||<:>||Kenshi and Hidetaka recommend Sazanami-Gothic. A stripped ttf is about 160KB. Still there is a [http://lists.debian.org/debian-boot/2006/02/msg00936.html question] if the kana and kanji glyphs are well mixed for Japanese.||

Intro

One of the main advantages of the graphical version of the installer is that it can support languages that use combining characters. We are currently using mainly TTF fonts, with the exception of Japanese and Chinese.

[http://fontconfig.org/wiki/ Fontconfig] is used for font management. See [http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=354044 #354044]

The following issues are open with respect to fonts:

  • Not all fonts needed for the graphical installer have yet been packaged into udebs (needed for inclusion in the official Etch release).
  • Translators should check if their languages are displayed correctly and check for missing characters. Special attention should be given to Indic languages, Arabic and Hebrew. Help from translators may be needed to decide the optimal font for the installer (taking into consideration both quality and space required). Please note that the screenshots are a little outdated currently.
  • Text inside GTK!?TreeView widgets is left-justified for RTL languages (see [http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=342239 #342239]), but that is fixed in the new DFB libs.

  • Ideal would be to have each character we need present in only one font. The current fonts have a lot of overlap which wastes space.
  • Display has to be good both when no language has yet been selected (language selection screen) and when it has.
  • Possibly font configuration can be improved. It would be very nice if we could influence which font is used for a specific language.

News

2006.03.03 - Davide Viti [http://lists.debian.org/debian-boot/2006/03/msg00073.html has sumarised] the status of the G-I before beta 2. The first issue is font related:

2006.02.24 - Davide Viti [http://lists.debian.org/debian-boot/2006/02/msg00915.html has created] [http://www.webalice.it/zinosat/g-i/mini_20060223.iso an iso image] and made [http://lists.debian.org/debian-boot/2006/02/msg00916.html sceenshots for all supported languages]. This iso image should fix all display issues [http://lists.debian.org/debian-boot/2006/02/msg00915.html except for Vietnamese] (due to missing glyphs in freefont). RTL languages' testers shall ignore the fact that in some cases the text is not alligned to the right edge (this is due to the usage of an old library but that will be changed as soon as possible).

Requirements for packaging fonts

  • Total amount of space used for fonts should be minimized.
  • Only font files that are actually used in the installer should be included.
  • Minimize overlap between fonts. It is possible to "strip" characters from a font (This is currently done in the developement versions with a fontforge script) We _really_ need to add characters as we need them, not to strip them from fonts. The fonts overlap too much - CJK fonts provide also codes for Latin scripts.

    Latin scripts are provided by most fonts, but they can be fixed with some [http://fontforge.sourceforge.net/scripting.html Fontforge scripting]

Davide Viti created [http://svn.debian.org/wsvn/d-i/trunk/scripts/g-i/strip_glyphs.pe?op=file&rev=0&sc=0 a fontforge script] that will allow stripping unnecessary glyphs or glyph ranges from the fonts. If needed, one can [http://svn.debian.org/wsvn/d-i/trunk/scripts/g-i/README.strip_glyphs?op=file&rev=0&sc=0 read the usage hints].

Information on how to play around with fonts is included in ["DebianInstallerGUIBuild"].


Fonts configuration for the G-I

Infos for language codes were fetched from:

Latin-1

Base characters (Latin) are taken from ttf-freefont (existing udeb). This font was selected because it looks quite good, has a character set that gives good support for a lot of languages and was already packaged as an udeb.

A negative aspect is that it also has partial/limited/ugly support for other scripts/languages and it would be better to only have specialized fonts for those. This is being resolved by "stripping" all glyphs from the font which we want to take from other fonts.

We also only really need the FreeSans* variants of freefont, and not everything that is currently in the udeb. ChristianPerrier has a nearly ready patch for this.

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

ca

Catalan

ttf-freefont

ttf-dejavu

ok, dejavu looks better

da

Danish

ttf-freefont

ttf-dejavu

ok, dejavu looks better

de

German

ttf-freefont

ttf-dejavu

ok, dejavu looks better

en

English

ttf-freefont

ttf-dejavu

ok, dejavu looks better

es

Spanish

ttf-freefont

ttf-dejavu

ok, dejavu looks better

eu

Basque

ttf-freefont

ttf-dejavu

ok, dejavu looks better

fi

Finnish

ttf-freefont

ttf-dejavu

ok, dejavu looks better

fr

French

ttf-freefont

ttf-dejavu

ok, dejavu looks better

ga

Irish

ttf-freefont

ttf-dejavu

ok, dejavu looks better

gl

Gallegan

ttf-freefont

ttf-dejavu

ok, dejavu looks better

is

Icelandic

ttf-freefont

ttf-dejavu

ok, dejavu looks better

it

Italian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

nb

Norwegian Bokmal

ttf-freefont

ttf-dejavu

ok, dejavu looks better

nl

Dutch

ttf-freefont

ttf-dejavu

ok, dejavu looks better

nn

Norwegian Nynorsk

ttf-freefont

ttf-dejavu

ok, dejavu looks better

pt

Portuguese

ttf-freefont

ttf-dejavu

ok, dejavu looks better

pt_BR

Portuguese (Brazil)

ttf-freefont

ttf-dejavu

ok, dejavu looks better

sq

Albanian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

sv

Swedish

ttf-freefont

ttf-dejavu

ok, dejavu looks better

Latin-other

These are also covered by ttf-freefont.

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

bs

Bosnian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

cs

Czech

ttf-freefont

ttf-dejavu

ok, dejavu looks better

cy

Welsh

ttf-freefont

ttf-dejavu

ok, dejavu looks better

eo

Esperanto

ttf-freefont

ttf-dejavu

ok, dejavu looks better

et

Estonian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

hu

Hungarian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

id

Indonesian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

ku

Kurdish

ttf-freefont

ttf-dejavu

ok, dejavu looks better

lv

Latvian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

lt

Lithuanian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

mg

Malagasy

ttf-freefont

ttf-dejavu

ok, dejavu looks better

pl

Polish

ttf-freefont

ttf-dejavu

ok, dejavu looks better

ro

Romanian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

hr

Croatian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

se

Sami

ttf-freefont

ttf-dejavu

ok, dejavu looks better

sk

Slovakian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

sl

Slovenian

ttf-freefont

ttf-dejavu

ok, dejavu looks better

tl

Tagalog

ttf-freefont

ttf-dejavu

ok, dejavu looks better

tr

Turkish

ttf-freefont

ttf-dejavu

ok, dejavu looks better

vi

Vietnamese

ttf-freefont

ttf-dejavu;[http://www.vn.net/staticpages/index.php/VietUniFont-HanNomSet ?HanNomSet],[http://www.vn.net/staticpages/index.php/VietUniFont-TriChlorSet ?TriChlorSet]

Some characters are missing in the current version of ttf-freefont; the missing glyphs have been added starting with the 20051206 CVS snapshot (see [http://lists.debian.org/debian-boot/2006/02/msg00738.html this] thread). See for instance the string for "Installing the base system" which is "Đang cài đặt hệ thống cơ sở"; dejavu state is unknown; the licenses for [http://www.vn.net/staticpages/index.php/VietUniFont-HanNomSet ?HanNomSet] and [http://www.vn.net/staticpages/index.php/VietUniFont-TriChlorSet ?TriChlorSet] have not been studied

wo

Wolof

ttf-freefont

ttf-dejavu

Display is correct with ttf-freefont, even the "ŋ" characters; dejavu state is unknown

xh

Xhosa

ttf-freefont

ttf-dejavu

ok; dejavu state is unknown

Non-Latin scripts

Georgian script

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

ka

Georgian

none

ttf-bpg-georgian-fonts

translator recommends BPG Glaho, currently there is no Georgian translation in D-I. ttf-freefont also has Georgian characters

Greek script

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

el

Greek

ttf-freefont

Greek is a little blurred but ok: see [http://lists.debian.org/debian-boot/2006/02/msg00662.html this] thread

Hebrew script

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

he

Hebrew

ttf-freefont

ttf-freefont was [http://lists.debian.org/debian-boot/2005/12/msg00511.html approved by Lior Kaplan].

Indic family scripts

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

bn

Bengali

MuktiNarrow

ttf-freefont

ttf-freefont still has issues, while Mukti displays correctly. New freefont have been succesfully tested and can be used after [http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=254113 #254113] will be fixed.

hi

Hindi

Gargi_1.7.ttf

Chandas, Lohit Hindi

Displays correctly after a proper stripping. New freefont have been succesfully tested and can be used after [http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=254113 #254113] will be fixed.

ne

Nepali

Gargi_1.7.ttf

Chandas, Lohit Hindi

Paras pradhan recommends using Garki_1.7.ttf. There are some problems with ranges covered by multiple fonts; Stripping is necessary.

pa_IN

Punjabi (Gumurkhi)

Saab

Lohit Punjabi, AnmolUni

Saab.ttf works fine; new freefont have been succesfully tested and can be used after [http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=254113 #254113] will be fixed. Lohit Fonts are most suitable for Linux System. [http://packages.debian.org/unstable/x11/ttf-punjabi-fonts Punjabi Fonts]; (soumyadip: well I was reading somewhere that AnmolUni was better) - this font should be (apparently) in ttf-punjabi-fonts, but is not currently (1:0.4.7)

km

Khmer

none(?)

ttf-khmeros

The [http://packages.debian.org/unstable/x11/ttf-khmeros ttf-khmeros] font should be used.

ml

Malayalam

none(?)

RachanaMedium.ttf ([http://packages.debian.org/unstable/x11/ttf-malayalam-fonts ttf-malayalam-fonts]), Racotf04.ttf

Vivek Varghese Cherian recommends using Racotf04.ttf (not packaged yet). Current font will be dropped in favor of Racotf04

Resources: #indlinux IRC channel on irc.freenode.net and eventually <debian-in-workers@lists.alioth.debian.org>

For Indic languages we currently include fonts from [http://packages.debian.org/unstable/x11/ttf-devanagari-fonts ttf-devanagari-fonts] and [http://packages.debian.org/unstable/x11/ttf-punjabi-fonts ttf-punjabi-fonts] (both maintained by Jaldhar H. Vyas <jaldhar@debian.org>)

  • Gargi_1.7.ttf (ttf-devanagari-fonts)
  • Saab.ttf (ttf-punjabi-fonts)

Bengali script (Brahmic family)

There was [http://lists.debian.org/debian-boot/2005/12/msg00338.html a report] stating that Bengali does not look right. [http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=343446 #343446] tries to analyze the problem.

Currently the installer uses Mukti Narrow, which should look right (recomended by Jamil Ahmed <itsjamil@gmail.com> and confirmed by Baishampayan Ghose <b.ghose@gnu.org.in>).

Mukti Narrow looks great, but there is some more stripping to be done (apparently) because this font is silently overridden by freefont (probably). (by silent I mean that "fc-list :lang:bn" does not list nothing other than Mukti, but the installer uses something else, anyway)

Update: #343446 confirms that ?MuktiNarrow should be used. Although freefont has support for Bengali, it seems that there are still some problems.

Devanagari script

[hi] Hindi

[ne] Nepali

Nepali uses the Devanagari script. Language is not currently supported by the Debian Installer (translations are not present).

Gurmukhi script

[pa_IN] Punjabi (Gumurkhi)

Punjabi Language is Written in Gurmukhi Script for Eastren Punjab (India).

Malayalam script

[ml] Malayalam

Malayalam has a script of his own. Vivek Varghese Cherian recommends using Racotf04.ttf. The suggestion to package it has been sent to Soumyadip Modak who maintains [http://packages.debian.org/unstable/x11/ttf-indic-fonts ttf-indic-fonts], the source package for [http://packages.debian.org/unstable/x11/ttf-malayalam-fonts ttf-malayalam-fonts].

Language is not currently supported by the Debian Installer (translations are not present).

South-East Asian scripts

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

km

Khmer

none(?)

ttf-khmeros

The [http://packages.debian.org/unstable/x11/ttf-khmeros ttf-khmeros] font should be used.

lo

Lao

none(?)

ttf-lao

A TTF font named "phetsarath" has been developed by the translation team. ChristianPerrier will likely package it. [http://bugs.debian.org/355088 The ttf-lao package is in NEW]

Khmer script

[km] Khmer

Language is not currently supported by the Debian Installer (translations are not present). Translation is under work and should appear soon.

Lao script

[lo] Lao

See "Status" column in the table above.

Arabic scripts

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

ar

Arabic

nazli

ae_AlMohanad and [http://lists.debian.org/debian-boot/2006/01/msg01119.html others]

Davide Viti has created [http://lists.debian.org/debian-boot/2006/02/msg00915.html an iso image] where he uses nazli for arabic, also [http://lists.debian.org/debian-boot/2006/02/msg00916.html sceenshots for all supported languages] were done an published

fa

Persian

nazli

[http://fpf.sourceforge.net/cms/index.php Free Persian Font] and [http://lists.debian.org/debian-boot/2006/01/msg01119.html others]

[http://fpf.sourceforge.net/cms/index.php Free Persian Font]'s italic [http://lists.debian.org/debian-boot/2006/01/msg01108.html looks ugly]. Tests should be done for "[http://lists.debian.org/debian-boot/2006/01/msg01119.html others]", if any of them are suggested by farsi speaking people.

nazli characters should be bigger (same for ae_AlMohanad in case it will be used again).

It seems that for arabic, nazli is the most fitted for arabic scripts.

[ar] Arabic

Resources: #arabeyes IRC channel on irc.freenode.net

Package: [http://packages.debian.org/unstable/debian-installer/ttf-arabeyes-udeb ttf-arabeyes-udeb] (maintained by Mohammed Adnène Trojette <adn+deb@diwi.org>)

The udeb includes "ae_AlMohanad.ttf"; as an alternative we've experimented a tarball containig "ae_Tholoth.ttf". "ae_AlMohanad.ttf" is said to produce better visual results.

See:

Needs to be checked now that ttf-freefont has been stripped.

  • There is some oddity going around in G-I as Arabic was not displayed when ttf-arabeyes was added along with ttf-punjabi-fonts, ?MuktiNarrow, ja, ko and zh fonts. This problem was not present when ttf-farsiweb is added instead of ttf-arabeyes.

[fa] Persian

Glyphs are shared with Arabic; "ae_AlMohanad.ttf" seem to have problems especially with ligatures; we need to see if "ae_Tholoth.ttf" works better. Roozbeh Pournader, upstream maintainer of [http://packages.debian.org/unstable/x11/ttf-farsiweb ttf-farsiweb] fonts, recommends using the nazli.ttf font from the [http://packages.debian.org/unstable/x11/ttf-farsiweb ttf-farsiweb] package. He discourages the use of ae_Tholoth.ttf. [http://packages.debian.org/unstable/x11/ttf-farsiweb ttf-farsiweb] is maintained by Clint Adams, but Arash Bijanzadeh, the Persian translation coordinator is ready to take it over as Clint wants to orphan the package. ChristianPerrier will help him in this process.

Chinese/Japanese/Korean scripts

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

ja

Japanese

ttf-cjk-compact-udeb

Kenshi and Hidetaka recommend Sazanami-Gothic. A stripped ttf is about 160KB. Still there is a [http://lists.debian.org/debian-boot/2006/02/msg00936.html question] if the kana and kanji glyphs are well mixed for Japanese.

ko

Korean

ttf-cjk-compact-udeb

[http://packages.debian.org/unstable/debian-installer/ttf-cjk-compact-udeb ttf-cjk-compact-udeb] is a custom made font set based on the necessary glyphs that are encountered in D-I. The Korean fonts are taken from UmDotum; We did have report from Sunjae Park <darehanl@gmail.com> stating that the [http://lists.debian.org/debian-boot/2005/11/msg01409.html ?UnDotum looked right]

zh_CN

Chinese (Simplified)

ttf-cjk-compact-udeb

ttf-comapct-fonts is a custom made font set based on the necessary glyphs that are encountered in D-I

zh_TW

Chinese (Traditional)

ttf-cjk-compact-udeb

ttf-comapct-fonts is a custom made font set based on the necessary glyphs that are encountered in D-I

For Chinese and Japanese [http://lists.debian.org/debian-boot/2005/12/msg00150.html we currently use the fonts prepared by Kenshi Muto] (ttf-compact-fonts). For an overview, see [http://people.debian.org/~kmuto/d-i/cjk/cjk.html this page].

We should make sure that none of the fonts below provides glyphs for other scripts than the ones which they were introduced (currently there are latin glyphs provided by them which take up space and harm us a lot for latin scripts).

Chinese and Japanese have an [http://en.wikipedia.org/wiki/Han_unification unification problem]. Unicode defines same codepoint against different glyphs between Japanese/Chinese(/Korean, but the Korean translation does not use any glyphs from this Unified Ideographs region so Korean is free from this issue). Furthermore Chinese font face and Japanese font face are different (do you think Helvetica and NewCentury are same? :-P ). So they don't like to use them together.

For Korean we now use [http://lists.debian.org/debian-boot/2005/12/msg00150.html ttf-compact-fonts], which has stripped UnDotum.ttf from ttf-unfonts (maintained by Changwoo Ryu <cwryu@debian.org>)

We previous included complete ttfs, but these took a lot of space:

  • UnDotum.ttf

  • UnDotumBold.ttf

We did have report stating that the UnDotum looked right: http://lists.debian.org/debian-boot/2005/11/msg01409.html

Cyrillic scripts

Lang Code

Language Name

Current Font

Other Fonts

Status / Observations

be

Belarusian

ttf-dejavu

bg

Bulgarian

ttf-dejavu

kk

Kazakh

ttf-dejavu

mk

Macedonian

ttf-dejavu

ru

Russian

ttf-dejavu

sr

Serbian

ttf-dejavu

uk

Ukrainian

ttf-dejavu

[http://dejavu.sourceforge.net/wiki/index.php/Main_Page ?DejaVu fonts] is based on Bitstream Vera and their recent releases have good cyrilic coverage. The [http://packages.debian.org/unstable/source/ttf-dejavu Debian package] has been updated to version 2.1. The package now creates a udeb since version 2.1-2, using the DejaVuSans fonts. If latin glyphs are not used from this font, then they should be stripped out.