Re: ligature usage - WAS: How do we find out what assigned code points aren't normally used in text?

From: Philippe Verdy <>
Date: Mon, 12 Sep 2011 19:55:22 +0200

Note: an old bug (signaled by me multiple times 2.5 years ago) is now being
corrected in Chrome since today: the Uniscribe part of the Webkit renderer
has a very critical bug that was finally isolated, it caused the whole
Windows desktop to become almost frozen or impossible to refresh on some
conditions, often forcing to reboot abruptly.

This old bug (appered in January 2009 with early beta versions of Chrome v4)
causes serious leakages of GDI resources (DC and font handles) in the
desktop window, within a very tight loop that sometimes never terminates.

It would be good if Chrome/Chromium and WebKit authors read this thread,
when they correct their Uniscribe support, in order to honor the language
markup or metadata in HTML5.


2011/9/12 James Cloos <>

> >>>>> "WL" == Werner LEMBERG <> writes:
> >> But "Dorfladen" is not ambiguous.
> WL> Yes, but some web browsers like Firefox automatically apply an `fl'
> WL> ligature...
> Only if the font does. (At least in the case of gecko-on-X11.)
> Ideally the text should be tagged as DE so that the app can call the
> opentype/graphite/whatever features for DE text rather than for generic
> latin (script) text.
> Failing that it would be useful to guess based on word lists, provided
> of course that doing do does not kill performance.
> -JimC
> --
> James Cloos <> OpenPGP: 1024D/ED7DAEA6
Received on Mon Sep 12 2011 - 12:58:25 CDT

This archive was generated by hypermail 2.2.0 : Mon Sep 12 2011 - 12:58:26 CDT