Devanagari in Mozilla: some issues
Sunil Bhadekar
I've run into a few bugs in the Mozilla browser/composer. I know you have
worked with the mozilla folks in the past, so thought you might be able to
get them across.
Bug #1: Mozilla doesn't justify Devanagari text correctly. For example, try
the following text in IE and Mozilla. IE displays it correctly justifying
the text on both ends.
<div style="text-align: justify;">
<p>नमस्ते, </p>
<p>हिन्दी न्यूज़ग्रूप पर दरख़्वास्त आई- भारत के राज्य और शहरों के बारे
में हिन्दी में जानकारी कहाँ मिलेगी। गूगल द्वारा खोज करने पर युनिकोडित
हिन्दी में ये जानकारी देने वाला कोई भी जालस्थल मुझे नहीं मिला। हिंदी में
ये जानकारी मुझे भारत सरकार के जालस्थलों पर PDF आवृत्ति में प्राप्त हुई।
अंग्रेज़ी में ये जानकारी देने वाले काफ़ी जालस्थल मिले । सोचा युनिकोड
में भी इसे उपलब्ध किया जाए.. </p>
<p>ये याद रखिये कि भारत के हर क्षेत्र की भाषा और बोली अलग होने के कारण
नीचे दिये गए नाम सिर्फ़ हिन्दी बोलने वाले ही प्रयोग करते हैं। हो सकता है
कि बाकी भाषाओं / क्षेत्रों में नाम अलग हों। </p>
</div>
Bug #2: Mozilla (and Composer) doesn't understand or display the
ZeroWidthJoiner and ZeroWidthNonJoiner combinations correctly:
1. क् + त = क्त (a glyph of kta)
2. क् + ZWJ + त = क्त (half-ka & ta)
3. क् + ZWNJ + त = क्त (ka-halant & ta)
Bug #3: Mozilla (and Composer) doesn't understand or display the Marathi
crescent-R sequence correctly:
र् + ZWJ + य = र्य (crescent-R & ya)
In Mozilla, the cresent-R is only displayed as an
r-halant-r-halant-consonant combo. For example: र्र्ह. (i.e. र् + र् + ह)
Bug #4: (Low Priority) After a syllable has been created using a Unicode
sequence, it should be the unit of display and cursor movement.
Additionally, when you select the text with a mouse, the raw Unicode letters
are displayed as you select the text creating movement in the text, whereas
the display should remain static as you select it. IE/Windows does both
these things correctly.
Bug #5: (Low Priority) IE/Windows displays Devanagari pages encoded in
Unicode UTF-8.
It also displays pages encoded in ISCII (x-iscii-de) by first mapping the
ISCII character table to Unicode (almost a one-to-one mapping) before using
the standard display routines. Could Mozilla support this?
If these could be fixed, it would be great! क्या आप मोज़िला में मिलने वाली
और कुछ खोटों को जानते हैं?
शुक्रिया,
सुनील
Alok Kumar
It's actually very easy to report bugs in mozilla via bugzilla. The
bug archive is online and searchable.
You didn't mention your mozilla version and the os on which you were
able to simulate them. There is a fine line (many times undemarcated?)
between the os's responsibility and the app's responsibility for
rendering devanagari.
> Bug #1: Mozilla doesn't justify Devanagari text correctly. For example, try
> the following text in IE and Mozilla. IE displays it correctly justifying
> the text on both ends.
>
> <div style="text-align: justify;">
> <p>नमस्ते, </p>
> <p>हिन्दी न्यूज़ग्रूप पर दरख़्वास्त आई- भारत के राज्य और शहरों के बारे
> में हिन्दी में जानकारी कहाँ मिलेगी। गूगल द्वारा खोज करने पर युनिकोडित
> हिन्दी में ये जानकारी देने वाला कोई भी जालस्थल मुझे नहीं मिला। हिंदी में
> ये जानकारी मुझे भारत सरकार के जालस्थलों पर PDF आवृत्ति में प्राप्त हुई।
> अंग्रेज़ी में ये जानकारी देने वाले काफ़ी जालस्थल मिले । सोचा युनिकोड
> में भी इसे उपलब्ध किया जाए.. </p>
> <p>ये याद रखिये कि भारत के हर क्षेत्र की भाषा और बोली अलग होने के कारण
> नीचे दिये गए नाम सिर्फ़ हिन्दी बोलने वाले ही प्रयोग करते हैं। हो सकता है
> कि बाकी भाषाओं / क्षेत्रों में नाम अलग हों। </p>
> </div>
(Quoted text may appear garbled, see original messasge.)
1. http://geocities.com/alkuma/2003/4/12/linuxrh71indixmoz13/bug1.jpg
2. http://geocities.com/alkuma/2003/4/12/win2kmoz13/bug1.jpg
3. http://geocities.com/alkuma/2003/4/12/xpwithlangpackie6/bug1.jpg
4. http://geocities.com/alkuma/2003/4/12/xpwithlangpackmoz11/bug1.jpg
The alignments seems to be the same (and correct?) in all four cases.
Case 2 doesn't display conjuncts and matras properly - that's a known
bug.
>
> Bug #2: Mozilla (and Composer) doesn't understand or display the
> ZeroWidthJoiner and ZeroWidthNonJoiner combinations correctly:
>
> 1. क् + त = क्त (a glyph of kta)
> 2. क् + ZWJ + त = क्त (half-ka & ta)
> 3. क् + ZWNJ + त = क्त (ka-halant & ta)
>
> Bug #3: Mozilla (and Composer) doesn't understand or display the Marathi
> crescent-R sequence correctly:
>
> र् + ZWJ + य = र्य (crescent-R & ya)
>
> In Mozilla, the cresent-R is only displayed as an
> r-halant-r-halant-consonant combo. For example: र्र्ह. (i.e. र् + र् + ह)
>
(Quoted text may appear garbled, see original messasge.)
1. http://geocities.com/alkuma/2003/4/12/linuxrh71indixmoz13/bug3.jpg
bug 2 - zwj is being displayed the way zwnj should be.
bug 3 doesn't show up here.
2. http://geocities.com/alkuma/2003/4/12/win2kmoz13/bug3.jpg
No conjuncts displayed anyway.
3. http://geocities.com/alkuma/2003/4/12/xpwithlangpackie6/bug3.jpg
Everything's perfect.
4. http://geocities.com/alkuma/2003/4/12/xpwithlangpackmoz11/bug3.jpg
bug 2 - zwj and zwnj combinations are not displayed.
bug 3 - r + y is not displayed properly but r + r + h is.
Summary: shows up fine on ie6 xp with lang pack. Mozilla had problems
though - if you ignore the win2kmoz13 case (which is not able to
display conjucts anyway), even on xp with lang pack, moz1.3 could not
display the kta with zwj & zwnj, not could it display rya. Should be
tested with moz1.3 on xp with langpack to see correct status.
> Bug #4: (Low Priority) After a syllable has been created using a Unicode
> sequence, it should be the unit of display and cursor movement.
> Additionally, when you select the text with a mouse, the raw Unicode letters
> are displayed as you select the text creating movement in the text, whereas
> the display should remain static as you select it. IE/Windows does both
> these things correctly.
Was able to simulate this on mozilla1.1 with xp lang pack as well as
on moz1.3 with linux rh7.1+indix.
Question: Is it a bug or a feature? :)
>
> Bug #5: (Low Priority) IE/Windows displays Devanagari pages encoded in
> Unicode UTF-8.
>
> It also displays pages encoded in ISCII (x-iscii-de) by first mapping the
> ISCII character table to Unicode (almost a one-to-one mapping) before using
> the standard display routines. Could Mozilla support this?
This would be a feature request. I'm not sure I'd like to *encourage*
this though.
Very few people actually know about the nuances of zwj and zwnj from
the Marathi angle, and you are one of them. That could be valuable to
the mozilla i18n team.
Also see http://bugzilla.mozilla.org/show_bug.cgi?id=166520
You could raise bugs/feature requests using
http://bugzilla.mozilla.org/enter_bug.cgi?format=guided , and mention
this thread/urls for simulation. You can also track their status from
http://bugzilla.mozilla.org/
The mozilla i18n team is planning a web page on CTL (complex text
layout), that should make it easier for users to report bugs --
particularly, to find out who to report it to - mozilla or the OS
vendor.
==
http://groups.yahoo.com/group/linux-bangalore-hindi/
can't see Hindi?
http://geocities.com/alkuma/seehindi.html
Alok Kumar
Surprisingly, http://posting.google.com is posting devanagari correctly today.
लगता है गूगल की गल बन गई।
Sunil Bhadekar
I couldn't access the geocities url's you had posted.
I'm using mozilla 1.3 on Windows 2000 with Indic enabled.
The fact that IE6 displays everything the way I expect probably means
the OS is fault-free?
- Sunil
Sunil Bhadekar
Take a look at this webpage (and the linked jpg image) which compares
the alignment. The first para is justified (should be flush on both
ends), the second is left-justified (should be flush to the left), the
third right-justified.
Mozilla displays the 2nd and 3rd correctly, but can't handle the 1st.
This is with Mozilla 1.3 on Windows 2000 with Indic enabled.
- Sunil
Sunil Bhadekar
Bug #1: Text Alignment
Take a look at this webpage (and the linked jpg image) which compares
the alignment.
http://members.tripod.com/~sunilbhadekar/justify-test.html
Alok Kumar
the os's.
> Bug #1: Text Alignment
>
> Take a look at this webpage (and the linked jpg image) which compares
> the alignment.
>
> http://members.tripod.com/~sunilbhadekar/justify-test.html
Oh yes, this is a bug. I've noticed this many times myself, but
thought it was because the author didn't key in the words properly.
Mozilla is trying to "stretch" the words to make the column flush on
both ends, but it's breaking the words in the process, whereas it
should be messing only with the spaces in between. I know of at least
3 webpages that display the same behavior.
http://stardataentry.com/hindi_book/home.html
http://www.indiapress.org/dharma/ (The Hindi text at the bottom)
http://iceprincess.livejournal.com/2002/08/07/ (Devanagari
conversation text)
Haven't looked at the html code for these pages though.
> I couldn't access the geocities url's you had posted.
Something to do with not having and index.html page in the directory.
I'll try to fix that, anyway those shots corroborate your findings
w.r.t bug 2&3.
What you could do:
1. Raise bug in bugzilla for bug1, the juxtaposed screenshot and html
text should suffice for simulation.
2. Raise bug in bugzilla for bug2 & bug3 (the test cases you mentioned
are kta plain, kta with zwj, kta with zwnj, r + ya, r + r + h, if
there are any other cases, you could add them as well)
3. Raise bug in bugzilla for bug4
4. Raise bug (feature request) for bug5, and then ask people to vote
for the bug.
If you don't feel like using bugzilla please let me know I'll do that.
If you do, please also post the bug id's so we can track the status
and provide more information to fixers if they need any.
Alok Kumar
> Alok,
>
> I couldn't access the geocities url's you had posted.
>
http://www.geocities.com/alkuma/2003/4/12/xpwithlangpackie6/bug3.jpg
http://www.geocities.com/alkuma/2003/4/12/win2kmoz13/bug3.jpg
http://www.geocities.com/alkuma/2003/4/12/linuxrh71indixmoz13/bug3.jpg
Kalidas
>
>http://members.tripod.com/~sunilbhadekar/justify-test.html
>
>The first para is justified (should be flush on both ends), the second
>is left-justified (should be flush to the left), the third
>right-justified.
>
This URL works very well with Mozilla 1.3a on Redhat Linux 8.0
Jai Hanuman
Alokji,
for windows users, takthi now supports posting to usenet via the
mail2news gateway. u can easily do
that if outlook express is you registered email program.
outlook express also allows you to post via hotmail
(similar to the SMTP relay).
"Alok Kumar" <alok....@mailjol.com> wrote in message
news:cba03d0.03041...@posting.google.com...
> > display conjucts anyway), even on xp with lang pack, moz1.3 could not
> c/moz1.3/moz1.1/
> Surprisingly, http://posting.google.com is posting devanagari correctly
today.
> लगता है गूगल की गल बन गई।
>
> > ==
> > http://groups.yahoo.com/group/linux-bangalore-hindi/
> > can't see Hindi?
> > http://geocities.com/alkuma/seehindi.html
==========================
free hindi unicode editor at
www.geocities.com/hanu_man_ji
Alok Kumar
> text should suffice for simulation.
http://bugzilla.mozilla.org/show_bug.cgi?id=202351
> 2. Raise bug in bugzilla for bug2 & bug3 (the test cases you mentioned
> are kta plain, kta with zwj, kta with zwnj, r + ya, r + r + h, if
> there are any other cases, you could add them as well)
http://bugzilla.mozilla.org/show_bug.cgi?id=202352
> 3. Raise bug in bugzilla for bug4
http://bugzilla.mozilla.org/show_bug.cgi?id=202354
Alok Kumar
We are looking for people to help us localize Mozilla to Indian languages.
Rediff has agreed to hire people for this and their objective is to
encourage the
penetration of the Internet in India. The programmers will work closely with
the IndLinux.org team to localize Mozilla and the software thus developed
will
be released in the public domain. This would be a good opportunity for those
passionate about Indian language computing to work full-time in this area.
The individuals selected will work out of Rediff's Mumbai office.
More details are below. Those interested are requested to write to us at:
ve...@indlinux.org and karu...@indlinux.org
Regards,
Venky
Mozilla Indianization
==============
1) Goal: Enable Mozilla browser to display & handle indian languages
properly, with proper font technology available. Localize Mozilla interface
also to indian languages.
2) Project Requirements: All work done should be contributed back to Mozilla
codebase, so that it becomes part of it. Solutions arrived at should make
into main tree, rather than be done as adhoc hacks / addons / patches.
i) i18n part (involving changing mozilla codebase) :
a - Should be able to render Indian languages encoded in Unicode/UTF-8
b - Use a standardized font technology - eg OpenType.
c - Support any encodings standardized for a language eg ISCII, TSCII,
KSCLP (KGP)
ii) l10n part (no code change - UI customization, translations of UI and
Help text):
a - Incorporate locale support (chrome level)
b - Translate Mozilla UI into indian languages
c - Making indianized mozilla theme - custom graphics, logos
3) Skill sets req
a - Knowlegde of at least two Indian languages - reading & writing, script
& language processing level 2.i & 2.ii.b
b - C, C++, XML, Java for 2.i
c - Graphics expertise for 2.ii.c
d - For translator - good vocabulary in English & at least one Indian
language, previous exp in translation also helpful.
e - Proj. mgmt skills, ability to work with volunteers remotely, coordinate
work through email/IM/group meetings.
f - Interest in localization and passion for having indian languages on
mozilla.
g - Work experience of 2-3 years in the IT industry
Current status:
Mozilla supports Unicode, so all indian languages represented in Unicode
work at unicode level - no proper rendering, if a suitable unicode font is
available. ie proper vowel sign positioning, conjuncts etc are not
avialable. Some work is on by Prabhat Hegde from Sun Microsystems on adding
devanagari support, basically by using Pango (www.pango.org) rendering
engine. He is exploring doing it using Opentype fonts. (Any one working for
this should work with Prabhat for
guidance & review, as he already has experience in Mozilla codebase )
Some hindi translation work has been done by a group based in Mumbai,
available at http://www.bttlindia.com/mozilla/ . Some translation work for
Tamil is also on. (These translations should be reused, instead of starting
from scratch)
Notes:
2.i has to be done in active participation with Mozilla developers & done at
those forums ( mozilla-i18n , mozilla-l10n )
2.ii.b is pure translation work, can be done in centralized or distributed
fashion.
2.i.c - Since ISCII, TSCII etc are not registered character sets, it may
not be possible to use them with Mozilla autodetection feature. Some
workaround could be suggested. ISCII text will basically be converted to
Unicode internally for display, enabled through use of 'iso8829-12'
reserved for Indic - ISO is waiting for us to settle to something :). (The
rationale to support ISCII is since ISCII chars will take 1byte, UTF-8 indic
characters take 3bytes, or 7 bytes when used as named
character references - The debate will pick up heat when indian languages
are fully supported )
-0-