Re: hyphenation in dillo

Hi Jorge,

On Wed, Aug 22, Jorge Arellano Cid wrote:
> Hi Sebastian,
> 
>   Sorry for the late reply, part cybernetic part human.
> As you noticed there was sometihng weird with the email:
> 
> <quote>
>    Delivery-date: Mon, 06 Aug 2012 20:58:14 +0200
>    Received: from localhost ([127.0.0.1] helo=auriga.wearlab.de)
>            by kbs62.informatik.uni-bremen.de with esmtp (Exim 4.69)
>            (envelope-from <[email protected]>)
>            id 1SySVI-00054b-C4; Mon, 06 Aug 2012 20:58:12 +0200
>    Received: from lilly.ping.de ([83.97.42.2])
>            by kbs62.informatik.uni-bremen.de with smtp (Exim 4.69)
>            (envelope-from <[email protected]>) id 1StgCZ-0003or-8d
>            for [email protected]; Tue, 24 Jul 2012 16:35:07 +0200
>    Received: (qmail 24854 invoked from network); 24 Jul 2012 14:34:46 -0000
>    Received: (ofmipd 188.109.158.229); 24 Jul 2012 14:34:24 -0000
>    Date: 24 Jul 2012 16:31:16 +0200
> </quote>
> 
>   It took near twelve days to reach me...

Yes, there were technical problems. Hopefully solved now.
 
> On Tue, Jul 24, 2012 at 04:31:16PM +0200, Sebastian Geerken wrote:
> > Hi!
> > 
> > At <http://flpsed.org/hgweb/dillo_hyphen/>, you'll find a version of
> > dillo which automatically hyphenates words, using the algorith by Frank
> > Liang (which is also used by TeX). It is still quite messy, but works
> > already rather stable. Some manual work is neccessary (from
> > dillo_hyphen):
> 
>   Good work!

Thanks. BTW, Johannes did some improvements on the actual hyphenation
to reduce memory consumtion; see <http://flpsed.org/hgweb/dillo_hyphen_mem/>.
 
>   I've downloaded/compiled and tested a bit: even in spanish!
> I like how it breaks paragraphs by squeezing some lines and moving
> words around (even without hyphenation). The hyphenating part also
> seems OK (I need to test more).

Yes, line breaking has been improved a bit, by optimizing per line.
May perhaps even be improved by applying TeX's "total fit" algorithm.

As for the hyphenation, Liang's approach leads to some errors
(including my surname, which is hyphenated as "Ge-er-ken", instead of
"Geer-ken"), but this can be dealt with by defining exceptions.

>   This looks very promising and makes me look forward for the merge
> day! So far, dillo has served me very well for reading articles (my
> font/size, my colors), and now there's the extra of better line/word
> breaking. This will surely also be welcomed by our users.  Very good.

Yes, I'd like to have this in the main repo rather soon, especially to
avoid merging problems with other works.

>   There're  glitches,  sure.  Just  in  case you don't yet have a
> reliable  test  case,  lwn.net  has  problems with screen redraw.
> Lefthand margin.

I know of a problem relating tables, but somehow was not able to
construct a test case. I found nothing special about lwn.net, but I'll
ask you to test again after the fix.

Sebastian