Re: [Freetype] discerning encodings in ttf

[Antoine Leca]

Hi Alexis,

Alexei Podtelezhnikov wrote:
> 
> So at the moment we have the following unanswered questions.
> What would be the way for ttf developers to set up their fonts so that
> it's _easy_ for users to use them in open source comminity?

Do you know the TrueType patents problem?
If yes, don't you think this impedes finding _any_ solution to the "easy"
use of TTF fonts in Open Source community, at least at short term?
This is not that I believe I (or David) can be sued by Apple (would a
big PR mistake), but it merely restricts what we can do formally...

Then, there is a second problem. TrueType fonts require quite much work
to be good than require Type 1 (because the curves are simpler, they
require more hints; and since the hinting is made using a programming
language, this ends much more difficult). As a result, there exists very
few if any good free (as per Open Source) TTF, and the Open Source
community is reluctant to use anything that is not free (and also fonts
have a long history of being pirated, which add to this problem).

I should add that the Open Source definition is a bit difficult to apply
to TTF, since Open Source requires that the text listing being the normal
way to modify sources, and fonts are likely to be modified rather with
graphics tools. But this is an aside.

> What would be the way for the users to set them up? Read on...
> 
> It's good that you brough up CJK fonts. Here is the most well-known
> kochi used by Debian and RedHat at least:
> http://www.on.ics.keio.ac.jp/~yasu/linux/fonts/
> They are beautiful in Japanese range, but they are horrible in Latin and
> Cyrillic ones. It's too much to ask from the developer of this _huge_
> set to be just as artistic in the ranges where he is simply not
> interested in. He did the proper OS/2 tables recently. That's enough.
> Now, how can users easily use them? No good tool for generating fonts.dir
> and fonts.scale exist. They all ignore OS/2 tables.

I shall say then the problem is with these tools.

In Windows world, Microsoft with a similar problem, did invent a way
to deal with mixed font in the rendering of a document, using the available
fonts, without any support from the document itself. For these reason
Microsoft did expand the support of Unicode ranges in the OS/2 table, BTW.
I know that Netscape is committed to use a similar scheme for Mozilla (and
do have big problems to handle the X11 situation).

What did Microsoft: they don't acted at GDI level (i.e., at Freetype
level).
Instead, they create a (new) layer between the font rendering engine (GDI)
and the application (for example, Internet Explorer); this new layer is
called Text Encoding I-don't-remember-what, but the real part of it is
the RichEdit control.
What I say is that X11 should have a similar "layer". Whether it is already
present in X11 architecture or should be added is outside my competence
regarding X11. But I hear that the X11 font model is currently being
remodeled, and I am sure it will take your point into account.

The bottom line, however, is that the answer does not belong to Freetype,
which should be kept at a lower layer.

 
> Bottom line is that the set up can only be done be an experience person

I disagree. It _should_ be done automatically.
Computers should _never_ be restricted, in any way, by any task "to be
done by an experienced person". If it appears to be so, then computers
should be instructed, to be able to act automatically. (That is not to
say that the automatism should always rule. At the contrary, I believe
there should have ways to disconnect them. But it should not be required
to always use the manual setup).

> and one has to be present in case we want to add other fonts to the
> directory. Nothing can be done automatically. I am complaining because
> I don't enjoy beautiful Japanese, I suffer from horrible Cyrillic here.

I get your problem...

> And I blame freetype for not being able to ignore Cyrillic which is
> properly marked out.

... but I completely disagree with your diagnostic. Freetype is just a
rendering engine. It is instructed to render glyphs, so it renders
glyphs. Which glyphs? The ones the application using Freetype ask for.
As a courtesy, Freetype allows access to the character mapping, which
give the passing from Unicode (or whatever) encoding to glyph number;
but it also gives access to the OS/2 table, which as you explain can be
used to "know" what for ranges are available. Now, if the application
using Freetype, do make use of the character mapping facility, but do
not use the 'OS/2' information, why on earth should Freetype be blamed?


> Well, not freetype itself, but whoever uses it.
> How do I have them ignored step by step in the present framework.
> 
> * So basically, a lot of people would like to see some HOWTO from freetype.

<STANDARD_ANSWER>
We gratefully accept any contribution. ;-)
</STANDARD_ANSWER>


> * Second question: who else but you I can complain to?

I believe X11 guys (perhaps someone can give good links here, I can't).

> * RFE: more advanced freetype utilities are needed bad

Refer to the standard answer just above...

 
> PS. I'll try to find time to add OS/2 reverence to ftdump.

Good.



Antoine