Re: [Devel] beating the patents - an idea

[David Turner]

Hello,

address@hidden wrote:
> Hi, hope this email ends up somewhere near the right place... -please
> forward it for me if it does not, as it just might be important.
>
> Don't worry replying to this email unless you would like to.
>
> I have come up with some suggestions/possible ideas for the
> freetype engine, perhaps they could solve some big problems for you:-
>
> * To get around the patent problem with true type font rendering,
> -would it help at all to provide your own "hint-files" to accompany
> many of the main popular ttf fonts? (arial, times new roman etc.) thus using
> your own database of good hinting adjustments at various common small sizes?
> The rendering engine would look for a corresponding "hint file" for the
> font it was using and read the list of hints from it for the size, angle
> etc. that it was trying to render.
>
Well, it's do-able, but the problem is to know how you're going to generate
these "external" hints. Nobody's going to do that manually, and an algorithmic
process is going to either be based on:

  A - auto-hinting (which we already do at run-time anyway)
  B - the native bytecodes, being interpreted

Unfortunately, the Apple patents basically cover two things when it comes
to interpretation:

  * using freedom/projection vectors to compute point displacements,
    which is critical to properly hint diagonals.

  * using "deltas" to move control points by explicit pixel distances
    for certain character sizes.

It's possible to put 'deltas' in external files (as far as I remember,
the patent claims are specific to the case where the displacement are
taken from the font file itself). However, there is little things we
can do to avoid the "vector" claims, because Claim 3 of US5155805
ambiguously says that any computation that results in the same final
vector is covered by the patent, so putting the information outside
of the font file would not help.

Unlike what Werner told you, I have actually thought about implementing
this technique quite a few years ago :-) But I've come to the conclusion
that we'd never be able to get rid of the patent menace by doing so.

So, I prefer to rely on auto-hinting for now, and possibly in the
use of "external" bitmaps in the near future. The trick is of course
being able to correctly "link" a bitmap to its outline font.

> In fact, would it break any patents to use your existing rendering engine
> (with a license) to generate the list of hints for you? you could make a
> custom version of it which outputted the hints to a file.
> (perhaps the patent only covers the formulae, not the actual values of the
> hints...)
>
Unfortunately, it seems to cover the values.. :-(

> You might even be able to provide the customized engine to those who wanted
> it to create their own hint files for their own fonts which they might have
> hinted themselves.
Nobody's going to use a custom hint format, believe me. Font designers don't
want to mess with such niche tools (they don't care much about Unix in general),
and computer hobbyist are generally poor font hinters...

Developing such a scheme would be a waste of time in my opinion. I believe
however that providing external bitmaps should be a lot easier to do (since
you can generate anyway you like, automatically or not).

Actually, we plan to support this through FT_Attach_File as discussed
recently in this mailing list.


Thanks for the idea,

- David Turner
- The FreeType Project  (www.freetype.org)