Re: Web font test cases: file linking vs. data URI embedding, CSS font stack subsets

Richard Fink wrote:

> Please explain what you mean by splitting, John.

As Vlad, explained, by splitting we mean serving glyphs from the same 
typeface as different fonts. This is an obfuscation method.

This is one of the obfuscation methods that Typekit are using -- or were 
when I discussed it with them at TypeCon. They were aware of the kerning 
issue, and that as more OpenType Layout becomes accessible in browsers 
this method will not remain viable for many fonts.


Subsetting is a different matter, and can be done intelligently so that 
OpenType Layout is preserved (by rewriting the Glyph ID references in 
the OTL tables) as is the case with WEFT. Font foundries wanting to ship 
subsetted WOFF or other font formats to web font licensees will need to 
figure out workflows to handle subsetting intelligently.

The most common kind of subsetting is likely to be by writing system and 
language needs. If a customer is only producing websites with English 
content, he isn't going to need glyphs for Cyrillic or Greek etc. Of 
course, he may opt for e.g. Latin 1 charset support, not a more strictly 
subsetted English-only character set, since foreign words requiring 
diacritics may occur in an English language context.

What I don't think we'll see much of is subsetting for specific web 
content, i.e. just the glyphs needed to render a particular page or 
pages. Web content is a lot more dynamic now than it was when EOT and 
WEFT were first developed with that kind of subsetting in mind.

John H.

Received on Tuesday, 13 October 2009 19:18:44 UTC