Re: Representative glyphs for combining kannada signs

From: Philippe Verdy (verdy_p@wanadoo.fr)
Date: Mon Mar 20 2006 - 20:00:42 CST

  • Next message: Andrew West: "Re: Representative glyphs for combining kannada signs"

    From: "Antoine Leca" <Antoine10646@Leca-Marti.org>
    > I do not know about XP here. However, since '05 2003sp1 also restricts
    > itself to the same scripts (i.e., no way to display the Syriac characters
    > from Estrangelo Edessa; now this DOES look like a bug, by the way), I guess
    > all the NT 5.x should behave in very similar ways
    > (LongHorn/Vista IS another beast, since at least I cannot launch LH's
    > charmap.exe here!)
    >
    >> Your screenshot seems to display glyphs in a font, not the codepoints
    >> to which they are assignedinternally.
    >
    > Playing around a bit, I can say at least 2000's/2003's Charmap only shows
    > the assigned codepoints (and only part of them, moreover, see above.)

    Charmap in XP does display Syriac characters with Estrangelo Edessa (and I've got no problem to render this script with that font or with other fonts, including in IE).

    But as Peter Constable just kindly explained, Charmap does use a simple text rendering API that depends on Uniscribe to render the glyphs mapped to codepoints in the font, it falls back to other fonts which are internally hardwired to some core fonts, if the script normally requires OT layout tables which are missing from the font.

    So the version of Uniscribe you use on Windows 2000, and the version of the Estrangelo Edessa font does matter for the results.

    What worries much more is that Uniscribe selects too much easily fallback fonts to display some scripts, even though the third party font has all the necessary glyphs and enough information in the OT layout tables to make that font usable for that script. And Uniscribe is used extensively and explicitly in IE for the line and page layout.

    However I still don't know if it uses a simple text API (similar to what Notepad does once it has determined how to break words and flow them on lines) to get the text bounding boxes andmake the final positioned text rendering : how does it determine which font to use from the "font-family:" CSS style, and how this coexists with Uniscribe fallback mechanisms.

    For me, the AUMSfont that provides glyphs for some scripts and pretends to support it, but does not have the OpentType layout tables that Uniscribe ABSOLUTELY WANTS, is causing much problems duetothe fallback mechanisms of Uniscribe. And IE cannot then select another font listed in the "font-family:" CSS style. Once AUMS has "said" to IE that it supports that script, it takes full control on that script and the restof the "font-family:" CSS style is ignored.

    This creates unsolvable issues for designing CSS stylesheets for multiscript web pages, because the relative order of fonts in "font-family:" is extremely important and also depends on the presence or absence of those fonts on the browser's system. A web author may create CSS stylesheets with list of candidate fonts for various systems (including known good third-party fonts), but how to order them is extremely tricky.

    For many users, extended Latin (IPA for example) only works with Arial Unicode MS, and it is often the only font they setup for the Latin script to get better coverage of the Latin script (the default font in IE setup for Latin are "Times New Roman" for "serif" and "Arial" for "sans-serif", but these defaults are hardwired in IE and are not changeable from the IE fonts options panel, where you can just select a common font for the two styles). A web author does not know if this AUMS font has been setupbythe user for the Latin script in IE, and if it is the case, it severely impacts the way other non-Latin texts are rendered.

    Things would have been much better if AUMS did NOT include any glyphs for scripts that require OT layout tables needed by Uniscribe (so excluding all Oria glyphs and mappings if Oriya OT feature tables are missing from the font). AUMS (1.00 or 1.01) is an quite good font for Latin, but it's preferable not to use it especially for the web. This is a hint given to Microsoft:
    (1) immediately provide a AUMS-like font (renaming it simply "Arial", and possible available through Office Update?) without these glyphs and partial support of scripts like Oriya, so that the font remains usable for the web and keeps good coverage of Latin. The removed scripts will give some place for other styles (bold and italic).
    (2) Then create a new version of AUMS for upgrades through Office Update with the necessary OT tables for the additional scripts.



    This archive was generated by hypermail 2.1.5 : Mon Mar 20 2006 - 20:04:20 CST