--- Begin Message ---
|
Subject: |
[OpenType] Re: TrueType adv.width scaling |
|
Date: |
Sat, 4 Mar 2006 06:31:00 -0800 |
OpenType list address: address@hidden
Not consistently, unfortunately... or not as pedantic as I would prefer it
to... The root of the problem appears to be a plurality of numerical formats
(16.16, 26.6, and 32.0), combined with innocent conversions (notice the plural
here) between those formats, entailing a plurality of conversion code paths:
16.16 -> 26.6 -> 32.0 vs 16.16 -> 32.0 vs ...
To cut a long story short: I tend to call this the "double rounding jeopardy."
First round a 16.16 to 26.6, then round the 26.6 to 32.0---as opposed to
rounding the 16.16 to 32.0 in a single step---and finally marvel at the
surprising deviations between the two code paths... I'll assume the legacy
thereof is that at the turn of the 1990s processor speed was at a premium,
while in the 21st century total cost of code maintenance is probably the more
dominant premium.
Yuri, if there's any combination of font/ppem size/Unicode # you're
particularly interested in, please don't hesitate to let me know and I'll try
to do my best (within what legal constraints I might be under) to answer the
question in excruciating detail.
All, 26.6 is a 32bit number with 26 binary digits (decimal numbers in range
-33,554,432 through 33,554,431) and 6 binary places (decimal fractions in range
1/2 through 1/64). These numbers can be represented and processed efficiently
by a computer of early 90ies vintage. Other formats are defined similarly,
according to the requirements of fractional precision, and any of these formats
may be perfectly valid in their respective context. The problem arises from the
repeated innocent conversion between one format and another; a bit like the
conversion from just about any European currency to Euros and *back*; someone's
going to loose in the process, and I bet it's not the banks...
Hope this helps
Beat
________________________________
From: address@hidden on behalf of Yuri Yarmola
Sent: Thu 2006-03-02 10:29
To: Multiple recipients of opentype
Subject: [OpenType] TrueType adv.width scaling
OpenType list address: address@hidden
I have one simple question: how TT machine scales advanced width when value
falls on the middle of the pixel at some PPM? It seems that it will produce
7 for 7.5 while CacheTT will produce 8.
No hinting is applied to right sidebearing at this PPM.
Best regards,
Yuri
Subscribe: address@hidden
Unsubscribe: address@hidden
Set list to inactive: address@hidden
Set list to active: address@hidden
Message mode: address@hidden
Digest mode: address@hidden
Subscribe: address@hidden
Unsubscribe: address@hidden
Set list to inactive: address@hidden
Set list to active: address@hidden
Message mode: address@hidden
Digest mode: address@hidden
--- End Message ---