[XeTeX] fontspec, xetex and Adobe Garamond & oddities

Jonathan Kew jonathan_kew at sil.org
Fri Oct 29 13:45:23 CEST 2004


On 28 Oct 2004, at 6:13 pm, Jonathan Kew wrote:

> On 28 Oct 2004, at 1:38 am, Will Robertson wrote:
>
>> BTW, Jonathan, is a bug with the OpenType library XeTeX? The 
>> "ordinals" OpenType feature only raises the very first character 
>> after a number, which makes it useless:
>>    %&program=xetex
>>    \font\agp="Adobe Garamond Pro:+ordn" at 20pt\agp
>>    1st 2nd 3rd 0th 6abcd
>>    \end
>
> I've looked into this, and it appears that the OpenType spec is 
> unclear regarding the precise format of the particular lookup type 
> used here (a chaining contextual substitution lookup). The Adobe Pro 
> fonts are based on one interpretation of the spec; the ICU library 
> implements an alternative reading. Some comments I've seen suggest 
> that OTLS/Uniscribe on Windows probably uses the same reading as ICU, 
> but I haven't yet been able to verify this.
>
> I can change the library XeTeX is using, and the ordinals in Adobe 
> Garamond Pro then work as intended; however, I'm concerned that there 
> may be fonts from other sources that assume the current 
> implementation, and these will break if I change it.

We're still seeking clarification of the OpenType spec for this type of 
lookup, but meanwhile, here's something Adobe's Thomas Phinney wrote a 
year or so ago on an OT mailing list:

> At Adobe, we have come to the conclusion that doing ordinals 
> contextually
> has extremely complex interactions with different languages, and that 
> it
> essentially isn't worth doing as a contextual feature. Instead, we've 
> gone
> to making it a simple substution in our newer fonts. This has some
> limitations/disadvantages (you need to turn it on only where you want 
> it
> instead of having it on all the time), but puts the control right in 
> the
> users' hands and ensures it's at least possible to get any desired
> behavior.
>
> At such a time as we revise older OpenType fonts, we will likely 
> revise our
> implementation of the 'ordn' feature as well to match the newer 
> approach.

This suggests that the contextual 'ordn' feature found in Garamond Pro 
(and Caslon Pro, and perhaps other fonts) is likely to be replaced with 
a non-contextual one at some point. I notice that some other Adobe Pro 
fonts (e.g., Minion Pro and Brioso Pro) implement 'ordn' as a simple 
substitution, so it's up to users to apply the feature only to the 
appropriate characters.

Given this situation, I don't think it's worth "fixing" the XeTeX 
engine to support the older Adobe fonts that are apparently going to be 
obsoleted, at the risk of possibly breaking other fonts that follow 
MS's interpretation of the spec.

JK



More information about the XeTeX mailing list