From: Asmus Freytag (asmusf@ix.netcom.com)
Date: Mon Sep 20 2004 - 15:26:44 CDT
At 11:50 AM 9/20/2004, Eric Muller wrote:
>But the real obstacle for a generative approach is QA: if as a font vendor
>you want to ensure some level of quality, then it is hard to avoid human
>work essentially proportional to the number of base+mark *combinations*
>you claim to support. If you don't *look* at a combination, at a number of
>sizes (to take into account rasterizing and hinting effects), in different
>contexts, then there is a chance that it does not happen as expected.
>
>I have no problem with people taking those chances or deciding their fonts
>are ok, or whatever. But I have a real problem if somebody else claims
>that *I* must take those chances, or that *I* must do an amount of work
>that is not justified by my commercial goals, or that *my* fonts are
>broken if I decided to not support some combination, or that *my* fonts
>are ok even if the result is below my standards.
That's why we are introducing named sequences. With a named sequence, it's
possible to define a repertoire that needs to be supported for a given set
of customers w/o requiring that each element of the repertoire be a
separate character. That allows the standard to be generative in nature,
but also allows customers and vendors to communicate which combinations
will be actively supported (with whatever font tuning and QA).
Just as no font supports *all* Unicode characters, no font supports *all*
combinations.
However, for named sequences to be useful, we must get buy-in from vendors
and users to use repertoires based on named sequences when specifying
fonts. In truetype there is rudimentary technology to refer to subsets
based on collections of character codes, but other than reading all the
Opentype tables, there is no method ti identify subsets that include
specific sequences.
A./
This archive was generated by hypermail 2.1.5 : Mon Sep 20 2004 - 15:28:04 CDT