Re: [Indlinux-group] Objectives, and deliverables for Sarai work at the IndLinux meet: May 16th, 17th

[Santhosh T. <san...@gm...>]

On Wed, May 13, 2009 at 10:39 AM, Gora Mohanty <go...@sa...> wrote:
> On Wed, 13 May 2009 09:44:06 +0530
> Santhosh Thottingal <san...@gm...> wrote:
>
>> Here is my list
>> Locales
>> 1. Locale independent sorting is not possible in GNU/Linux now. ie,,
>> You get different output if you sort Malayalam text in Tamil Locale or
>> Japan locale.
> [...]
>
> This was discussed extensively at FOSS.in, and the work
> done by Rahul, Pravin, et al, addresses this issue.

No.  I am not talking about that.  I am talking about the problem when
LC_COLLATE is defined in xx_IN and not in iso14651_t1_common file.
When sorting a language xx_IN in yy_IN locale collation rules of xx_IN
not taken and it defaults to C.
Just try sorting a file with more than 3 indian language using sort
command in en_US locale.  You will get a completely wrong output.

[I had confirmed this problem with Pravin Satpute]


>> Identify the problem by discussing with others, report bug.
>> I have a patch for fixing this in Tamil locale. Testing  it,
>> submitting to glibc.
>
> The point is not to fix it locale by locale, but to have
> a single sorting table that is simply included into each
> locale.
>

We define collation rules in iso14651_t1_common file with each locale
referring it. But ta_IN is not using common locale definitions and use
its own LC_COLLATE section  and make other locales fail.

Please refer ta_IN LC_COLLATE section of your distro.

>> 2. Spellcheck
> [...]
>> As of now, aspell or hunspell will work only in GNOME. In KDE ,
>> spellchecker will not work for Indian languages.
> [...]
>
> Er, what? Don't know about Hunspell, but aspell has worked in
> KDE for donkey's years. Just checked again, and it does work
> at least in kate, and kword, once the spell-checking dictionary
> is set.


I am talking about KDE 4.2. From 4.0,  KDE uses Sonnet as a library
for interfacing with various spell check backends.
While doing word boundary detection, it uses Qt's isLetter, isAlpha
functions. Those functions are buggy for Indian languages. Recently we
had fixed this problem in glibc. Since QT uses its own unicode
character database, this basic functions need to be fixed to get
spellchecker working in KDE(>4.0)
Refer http://santhoshtr.livejournal.com/14738.html for more details.
Spellcheck is just one feature that got affected by this QT bug .
And I came to know that even the latest Qt version also fails to do
word boundary detection

Thanks
Santhosh