- From: Sergey Malkin <sergeym@microsoft.com>
- Date: Tue, 17 Jul 2012 23:17:13 +0000
- To: John Daggett <jdaggett@mozilla.com>, www-style list <www-style@w3.org>
- CC: "www-font@w3.org" <www-font@w3.org>
My test was long list of simple paragraphs (Adam's email from this thread, in fact). Line width about 50-60 characters, 5-6 lines per paragraph, total file size was about 10Mb. I tried Times New Roman, Calibri and Algerian. Algerian was around the same speed as IE, so I assume this includes "frame construction". I was not able to precisely determine what constitutes end of formatting, so I defined it as when processor load went from 100% back to normal. If there is something that I can use to get better measurements, let me know. Again, I don't say it is not possible to optimize, but reality now is that having OpenType features enabled by default gets performance hit that we are uncomfortable with. Nothing prevents particular browser on particular platform to interpret 'auto' as 'turn default features on'. Having "good typography" doesn't stop with OpenType features, browser can add other goodies like optimal paragraph layout or even full hZ algorithm. This area is where browsers can and should differentiate. Rob wrote: > If CSS 'letter-spacing' is greater than zero, we disable discretionary ligatures, because it seems safe to assume that matches the author's intent. That behavior could (and I think should) be written into the spec. This is exactly behavior that should not be in the spec. Why would you turn off discretionary ligatures, but not standard? How do you know that some stylistic set features, e.g. cursive connections, are not broken by letter-spacing? Answer absolutely depends on specific font and can't be controlled automatically. If you are on that level of control, features should be turned off in the stylesheet explicitly. Thanks, Sergey
Received on Tuesday, 17 July 2012 23:17:47 UTC