Contrast etc.

Eben Sorkin eben at eyebytes.com
Fri Sep 21 19:01:04 BST 2007


I am a type designer and I want to suggest a different path. I do  
admit that this isn't going to be helpful immediately but is instead  
a long term kind of idea.

The problem that Linux font rendering has as far as I can see - which  
may not be far enough I will admit readily - is that fonts designed  
for one rendering scheme are being forced into another one. It's a  
square peg round hole thing. It may work pretty well - in the sense  
that it is usable - but it's probably never going to be really really  
good. And even fonts designed for Linux are still using essentially  
the square peg as their frame of reference.

I suggest that you *may* ( not should ) want to accept a specific  
rendering scheme and instead of working to make existing fonts look  
less bad - focus on developing fonts that take advantage of and are  
designed for that specific rendering scheme.

It's a radical suggestion. I admit that. But there are some great  
reasons to consider it.

One reason: that is actually what MS did with clear type, and true  
type - they came up with some theoretically better type-tech and then  
designed type specifically for it. I am not holding MS up as some  
kind of apogee of type nirvana, but I am saying that that method of  
development is likely to yield more satisfying results.

The natural question that will come up is what about sizes/hinting  
which is something I have not adequitely addressed in the comments  
above.

Rather than suggest that you come up with your own hinting scheme I  
think that the best answer is that you may want to build an OS & font  
function that automatically supports & swaps size specific outlines  
rather than developing alternative hinting models.

For those of you saying 'what?' or 'huh?' at this point; Sorry. I  
explain more fully: When hinting is used it distorts a core form into  
a new one for screen. The core shape is used for printing and for  
large sizes. What I am suggesting has it's roots in the fine  
tradition of Metal type - size specific outlines. With size-specific  
outlines you can tune the outline to the ppem or pixels in a direct  
rather than indirect manner.

It means writing a complex driver to swap forms on the fly - but also  
means that type designers can use their eyes to see what looks good  
on the OS in an immediate manner rather than after a tedious round of  
hinting - which may or may not be interpreted correctly or even be used.

It's a startlingly simple, pleasant & distributable approach. It  
isn't my idea though.

What I am getting at is explained by the originator of the idea,  
David Berlow; here:

http://www.rogerblack.com/blog/screen_fonts_history



More information about the ubuntu-devel mailing list