Re: GUADEC Results

[Havoc Pennington <hp redhat com>]

Jesper Skov <jskov redhat com> writes: 
> Actually, having discussed it a (little) bit at GUADEC, I think, long
> term, we'll want to use Pango. There's simply too few developers on
> AbiWord do allow us to recreate all the good stuff in Pango from
> scratch - not to mention that it would be silly to do so.
> 
> I've had an ultra quick look at the layout code, and I think what we
> want is pluggable Pango layout modules, just like there's already
> language modules.  I think the layout engine working with shapes was
> even on the Pango wish list - that's what we need. And we'll probably
> also put some effort into optimizing algorithms and code since doing
> full-page text layout is going to be way more expensive than "just"
> a little text used in widgets.
> 
> We haven't looked at the details, but Pango does a lot of the stuff
> that the AbiWord backend already does (there's a surprise!). Our
> largest hindrance is likely to be Pango portability - we'll need Pango
> on all our supported platforms (Win, QNX, Mac, Unix) before we can
> switch. That will be part of the job of converting AbiWord to use
> Pango - but hopefully there'll be non-AbiWord hackers willing to help
> on that.
> 
> This is all well into the future, but I'm sure we'll get it sorted out
> eventually. I'm personally 95% sure Pango is the way forward for
> AbiWord...
> 

I agree with that, my point was that you may want to use the Pango
itemizer and shaper, and e.g. text boundary algorithm, but not
necessarily PangoLayout. PangoLayout is a special-purpose layout
engine for the simple case, it's not intended to be comprehensive.
I'm not sure to what degree it makes sense to grow PangoLayout; it
needs to remain relatively small/fast so it can be used in text
labels, etc. So it could make sense to refactor bits of PangoLayout
into a separate file, and do a more industrial-strength layout engine
using those. 

Anyhow, all that's just implementation details. I think making
GLib/Pango as portable as AbiWord would be a nice step forward for
GLib itself, and no doubt using that layer XP will make some things a
lot easier for AbiWord.

Havoc