From: Philippe Verdy (verdy_p@wanadoo.fr)
Date: Thu Mar 16 2006 - 21:52:23 CST
From: "Antoine Leca" <Antoine10646@Leca-Marti.org>
>> This causes lots of confusion and errors in many fonts
>
> Of course, if font designers ONLY base thir work on TUS, the result is
> unlikely to be of ultimate quality.
> However, I do no believe this is common behaviour.
>
>> (as well as interoperability problems, when documents are created
>> with these fonts, notably for the web).
>
> Please elaborate, I do not see the point here.
>
>
>> The problem is notable even with core fonts provided with Windows, or
>> rendered with Uniscribe or in Internet Explorer, even when the fonts
>> are correct (the most significant error is the case of the vowel sign
>> I which is not displayed before the base letter in Bengali or Oriya
>> or Gurmukhi).
>
> Either the font is correct, and the problem does not show up, which is BTW
> my relation of the case you mentionned; in fact, for Gurmukhi I know quite
> well, every version of Uniscribe I know of, starting with the first beta I
> saw in 1998, deals correctly with Gurmukhi sihari, provided you use it with
> a properly encoded font (and early betas did have a different scheme for the
> features, also available with plain IE5).
>
> Or the fonts is plain incorrect. Correct placement of the I matra is a very
> basic mandatory feature.
Just try to view this French Wikipedia page [[fr:Table des caractères_Unicode (0000-0FFF)]] that shows the Unicode charts blocks in plain HTML (this page displays the first indic script blocks with code points under U+1000). Look at the places where a bold "Note :" indicates various rendering bugs found most often in Internet Explorer or in Firefox.
http://fr.wikipedia.org/wiki/Table_des_caract%C3%A8res_Unicode_%280000-0FFF%29
You'll see that "some" browsers still do not correctly handle the needed reordering of vowels before display and this includes Internet Explorer (IE6 or even in latest beta of IE7 on Windows XP with the latest patches installed, and with Office 2003 installed with additional fonts and most up-to-date Uniscribe engine). And if you copy/paste it from Internet Explorer to MS Office Word or Excel documents, where Uniscribe is used, things don't go really better (the results are different and show other bugs for other code blocks.) Now try to print them, and the preview and print renderer seems to use another algorithm with *other* bugs...
Things are much better in Mozilla Firefox, which supports much more scripts than Internet Explorer on running exactly on the same system and the same installed fonts (and the best font settings as possible specified identically on both browsers) but even Firefox has its own bugs (including bugs in the way it incorrectly remaps the codepoints to other codepoints to emulate missing glyphs by others found in some fonts; in some cases the tables it uses are completely wrong or were created based on Beta versions of Unicode, such as the Coptic alphabet block that it remaps very incorrectly).
Note that this table doesn't use a generic dotted circle to display combining marks, but explicitly chooses a valid character in the same script (indicated explicitly in the text before each table) used as the base for combining marks of each block. Such page alone can be a good and fast test page to see the effective and correct support of Unicode in browsers, and what to look for if there are rendering errors.
The mandatory placement of all matras placed before the consonnant is signaled as a serious bug, and this is what I call an "interoperability problem" because it forces authors to change the local document encoding specially for some browsers or OSes that only use a visual ordering for those scripts, just to produce the desired order at reading (and yes this category includes Internet Explorer, Windows itself, and Microsoft Office applications).
This archive was generated by hypermail 2.1.5 : Thu Mar 16 2006 - 22:12:05 CST