[OpenFontLibrary] use of (c) typefaces

[Ed Trager]

Hi, Alexandre,

On Fri, May 8, 2009 at 10:17 AM, Alexandre Prokoudine
<alexandre.prokoudine at gmail.com> wrote:
> On Fri, May 8, 2009 at 6:15 PM, Ed Trager wrote:
>
>> But I assume that the problem you are really trying to address is one
>> of people copying glyph outlines into a new font that they claim to be
>> their own?
>
> Yes
>
>>  For that kind of situation, one would, I assume, have to
>> try to match glyph outlines against some searchable database of glyph
>> outlines ...
>
> Exactly what I say :)
>
>> sounds hard to do ...
>
> But it works

Does it?  If someone was going to copy glyphs from some other font,
they might either intentionally --or indeed completely
*unintentionally*-- make changes to the glyph outlines.  Visually the
glyphs would still look very close to the originals, but the actual
points and curves might be sufficiently different from the original to
avoid detection by most simple matching algorithms.  For example,
instead of copying glyphs electronically, someone might scan printed
pages at a reasonable degree of resolution, then use bitmap tracing to
re-vectorize the glyphs.  Indeed, for legitimate revivals of old
printed typefaces that are in public domain, a number of useful tools
and scripts are available exactly for the purpose of reviving old
typefaces as modern digital "revivals".  Any "revival" from scanned
images will certainly result in unique outline vectorizations that
will always differ from the originals (in the case where the
"originals" were in fact produced from digital type).

So in fact it now seems to me that having a database of *outlines*
will be useless.  A better approach would be to create a database of
*bitmaps* of the glyphs at some sufficiently high pre-defined
resolution.  Then, to test a suspicious font, one would in fact
rasterize the glyphs to a set of bitmaps at the same pre-defined
resolution, then overlay and subtract one bitmap from the other (i.e.,
do some sort of pixel-aligned XOR operation on the two bitmaps) and
see what was left over.  If nothing or very little was left over, that
could be used to flag a font for review by a human.  Something like
this would probably work ...

Is this the idea you had, or some other idea?

>
> Alexandre
>