Has anyone succesfully installed squeak on RH 8.0?

Jim.Gettys at hp.com Jim.Gettys at hp.com
Fri Jan 10 20:45:23 UTC 2003


> From: Yoshiki.Ohshima at acm.org
> Date: Fri, 10 Jan 2003 11:36:29 -0800
> To: squeak-dev at lists.squeakfoundation.org
> Subject: Re: Has anyone succesfully installed squeak on RH 8.0?
> -----
>   Hello, Jim,
>
> > Recent X servers have the X Render extension: this adds the ability to
> > cache glyphs in the X server and paint them (alpha composited, with
> > Porter-Duff sematics), rather than dealing with
> > the old, broken, core X fonts (in the new model, fonts are entirely
> > client side objects); we have a library to uses Freetype to do the
> rendering,
> > and which also does *very* nice subpixel decimation for flat panels (Xft2);
> > the results of this are cached at the X server and then painted.
>
>   Oh, yes.  Several years ago, the X-TT server was getting popular,
> but if I remember correctly, it was a server side solution and just
> provides additional XLFD handing, right?

Yes, stay away from that horrid non-solution.

Possibly the biggest architectural mistake we made with X was
making fonts server side resources.

You might be interested in fontconfig for Linux (www.fontconfig.org).
This provides finally a unified font lookup mechanism that does
CSS2 style subsitution, and deals with solving the problem:

"What set of fonts does my application need that covers the following
languages?"

Fontconfig will tell you which fonts answer that question;
probably better than anything on Windows or Mac, and scales to
thousands of fonts.

Fontconfig is now in RH8, and Gnome 2.2 and KDE 3 are using it.

And it is not X specific at all, finally making it possible to
get on paper what you have on your screen for text...


>
> > But any other font rendering technology can be used, and the results
> > cached at the X server for fast painting.  Recent network tests show roughly
> > comparable bandwidth usage to the old core font mechanism, even though the
> > glyphs have to go over the wire.  Certainly much faster than pushing pixels
> > all the time (though Xft2 will do that as well, so that this all can be used
> > against old X servers).  The intent has been to get identical results
> > to the frame buffer whether or not X Render is present.
>
>   This may be over generalization, but most of people runs Squeak (an
> X client) and the X server on the same computer.  The less network
> usage is of course better, but it is not that important for this local
> configuration.

True.

>
> > Whether is solves anything for Squeak given the portability goals,
> > is a different question.
>
>   Yes.  You wrote "the X port needs work to take advantage of new X
> technologies...", but I would imagine that platform agnostic approach
> is preferable if possible.  Of course you can always hack the Squeak
> code because the sourse is available, but for mainstream, Henrik's
> FreeType plugin is nicer because it doesn't require the X, and pure
> Squeak solution would be better (if the rendering quality matches the
> FT's result) because it doesn't require anything.
>
>   TrueTypeTextStyle isn't going to make it into the mainstream image,
> I think, but it is on SqueakMap.  You can use it as the "system font"
> if you want.
>

Even if you ignore Render, Squeak may want to steal code from Xft2
(which uses Freetype for its rendering).

Xft2 however, implements (a typical geekish name) "subpixel decimation",
which gives better results in font rendering for flat panels
that what goes under the names "Cleartype" (Microsoft) or Cooltype (Adobe).
It does not use the typical signal processing approaches, but leverages
the fact that the eye is a strong edge detector, so characters should be
rendered more strongly to pixel boundaries.  Essentially, only pixels that
would be jaggy get antialiased, and use the flat panel's RGB subpixels
to improve the character's sharpness when doing diagonal strokes.
We believe (if truetype hinting is turned on) we have something here
better than what you see elsewhere for text rendering.

                                     - Jim

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



More information about the Squeak-dev mailing list