Re: What backend should be the default now on -nix?

[Nicolas Roard]

Le 26 oct. 04, à 23:35, Riccardo a écrit :

> Hello,
>
> On Sunday, October 24, 2004, at 07:22 PM, Fred Kiefer wrote:
>
> > don't you think we should exchange a few arguments on this before we 
> > start to vote? There are a few countries around were you may see the 
> > results of uninforemed voting :-)
> >
> > Here my comments on the different backends:
> >
> > cairo - great potential, when cairo itself is available everywhere, 
> > but at the moment does not display images properly and is to slow at 
> > the moment.
> very experimental.

to say the least. But it will probably be the "future" ...

> > art - most functional complete of them all, but requires additional 
> > libraries and fonts.
> good for eye-candy, anti-alias. I would say preferred backend for 
> demo-cd's and so. Few fonts. Has the advantage of using the original's 
> host fonts when exporting display (well it can be an 
> advantage/drawback depending on what you do). Has potential Shm 
> problems and is on some computers slower than xlib.

But faster on most of computers ..
And I think the graphic operations are better supported. For example, 
gradients. And rotations, clipped path, etc., although for theses it's 
perhaps also the case on xlib
now, it's been a certain time since I last tried it.
So I disagree a bit with you -- it's not only for "eye-candy" -- 
functionalities are better. And while you only seem to notice the AA 
fonts, it's not only the fonts that
are AA, but the whole display. Frankly, it completely change the way 
things look (from bad to good looking). (plus, it's possible to use 
non-AA font on backart, no ?

> exports worse (but works with shm disabled) than xlib.

Disagree here; Exports is only "worse" in the sense that you sometimes 
need to type a default by hand (but -- it works very well, and it's not 
something really complex
to do. And I think people using apps on remote are capable of typing a 
default line; and remote isn't exactly the most common use case)

For me, the only real problem with backart, is the fact that it uses 
its own font system. I personally like it and found it better; but I 
understand very well that it could be
held against it -- and in a non "gnustep os" (ie, not something 
dedicated to gnustep... that is... everywhere at the moment..) that's 
probably a problem :-)
The solution would be ideally to use fontconfig in addition to the 
nfont system...

> > xlib - works almost everywhere
> fast on many systems. portable. little dependencies. exports 
> excellently. has excellent non-aa fonts on system that have them 
> (quite usual). I like it (ok, ok, that's  not a big argument)
> But it is plagued by bugs! The font panel on most of my systems is 
> screwed up, showing fonts badlzy, not recognizing them all and 
> displaying a "Nil" font...
> I was told that some image operations are also unsupported... and it 
> is little unmaintained.
>
> It does not dither on 8bit displays. I don't know if this could be 
> implemented or not...

That's why people advocate the use of back-art ...

> > xdps - currently unsupported
> would be morally nice.. think about original NeXT... On systems with 
> dps (irix, solaris... openstep...) it would give excellent font speed, 
> incredible operations on objects, colors, fonts, rotations, 
> scalings... and excellent dithering.
> Most of the systems with dps are not supported by gnustep currently :)
> Also I don't know if there is a ocmparable "ghostscript" equivalent.

I agree that it would be nice in theory. In practice, I doubt that the 
work is worth it (the windows GDI backend, if any, is a much more 
important target for the project, imho)
If somebody really wants to use a dps system, he could works on it...

> > From that list I would say that xlib should be the default as it will 
> > run in all environments where any of the others will run. We could 
> > argue about the case, where we can proof inside of configure, that 
> > all the pre-requesits of art are fulfilled (e.g. all the libraries 
> > and fonts are there). In that case it would be fine for me to have 
> > art as the default. It wont help anybody to require loads of 
> > additional software just to get our backend running. Up to now the 
> > GNUstep approach was to use whatever is there and if something is 
> > missing rather leave out some functionality instead of not working at 
> > all.
>
> I quite agree.

The only addition to backart are 1) freetype 2) nfonts.
Freetype is generally already present on linux (and probably bsd) 
systems -- I don't think it's a problem. For nfonts, it's not really 
complex to use and to install.
What we can (should) do is provides an easy way of installing nfonts 
(perhaps provides the freefont directly in the tgz, something like 
that..)
And ideally work on adding fontconfig to it ?

But apart from some cases, most of the time, backart isn't complex to 
install and use -- and it then provides a better experience than xlib. 
That's why I think we should use it as default on X11, with xlib as a 
fallback. I think it will be better than xlib as default and backart as 
a fallback -- I think it's better to first try to provide the best 
experience if that doesn't work provides something else.

--
Nicolas Roard
"Any sufficiently advanced technology is indistinguishable from magic."
 -Arthur C. Clarke