> Sender: squeak-dev-bounces at lists.squeakfoundation.org
> From: "Andreas Raab" <andreas.raab at gmx.de>
> Date: Wed, 26 Feb 2003 10:20:49 +0100
> To: "'The general-purpose Squeak developers list'"
> <squeak-dev at lists.squeakfoundation.org>
> Subject: RE: Beta test Bitstream Vera fonts available.
> -----
> Hi Jim,
> 
> > There are no delta hints in Vera: they will not generate good
> > bitmaps at small sizes.  They are intended for use primarily
> > anti-aliased as we now do on Linux; antialiased, they do very
> > well indeed at small sizes.
> 
> Yup, that's what I thought.
> 
> > For some samples of where things are at using all open source
> > technology,
> > see: http://zap.crl.dec.com/Screenshot-4.png and
> > http://zap.crl.dec.com/Screenshot3.png.  This is best looked at
> > on a flat panel with typical horizontal rgb subpixel order.
> > Most things on those two shots are in one face of Vera or another....
> 
> Looks very impressive. In particular the small fonts look not washed out at
> all and appear very readable at small sizes.
> 
> > There will be some additional work done on hinting over
> > the next few weeks. You can see in the waterfall display
> > a few characters need some additional TLC, as also you observed...
> 
> *Grin* Compared to what I sent these look _perfect_ ;-)

Glad you think so.  For the record, note that I have the TrueType hinter 
enabled on my machine in Freetype; dunno how you have it set.

The latest freetype's autohinter (used if TrueType possibly
patented algorithms are not enabled) continues to improve btw....  
What you get out of the box on Linux depends on the Linux
distribution you use.

> 
> > Xft2 use Freetype for the basic rasterization; but that is
> > just the start. Xft2 adjusts the hinting to keep things on
> > pixel boundaries much more than Microsoft Cleartype does
> > (which appears to at most adjust to subpixel bounadaries);
> > we believe this is better, and I think the above shot of
> > my desktop bears this out.
> 
> Absolutely agree. The quality is amazing. So Xft2 hacks the hinting?! Sounds
> somewhat scary ;-)
> 

Yes, Keith Packard can be a bit scary at times... :-)  There is also one 
additional idea that could be tried; rather than keeping the character's 
strokes on pixel boundaries, keep a triad of subpixels together; that 
might allow better positioning of the glyphs on the line, but then the 
green subpixel would not necessarily be at the center (your eye is most 
sensitive to green).  Dunno if Keith will ever try that one out.  This 
all interacts in mystic ways with the human brain's signal processing 
system. Neurons are not the same as a DSP, and the only thing to do
is to implement and see if you like it better.  The human system
has *really* sensitive edge detectors...  And flat panels allow
you to get *MUCH* sharper edges; exploiting that is a major feature...

Basically, you only get AA on pixels that would otherwise have
ended up involved in jaggies...

> > The resulting glyphs are imaged at 3 times the horizontal
> > or vertical resolution (depending on flatpanel subpixel order
> > and orientation), and the results used to form a glyph which
> > is alpha composited to the screen.  This process Keith Packard
> > calls subpixel decimation.  So we end up a bit less faithful to
> > the font outlines, but less fuzzy than Cleartype (which is sharper
> > than not doing subpixel stuff is in the first place).
> 
> How's the quality for non-LCD screens?! Plain old CRTs for example?
> 

Fine as far as I can tell.  On plain old CRT's, you skip the
subpixel stuff, of course, and CRT's add "natural" fuzz whether you like
it or not.
                           - Jim



--
Jim Gettys
Cambridge Research Laboratory
HP Labs, Hewlett-Packard Company
Jim.Gettys at hp.com