[luatex] LuaTeX file almost 2.5x bigger

[Hans Hagen]

On 28-7-2010 12:55, Will Robertson wrote:
> On 2010-07-26 06:04:19 +0930, Hans Hagen <pragma at wxs.nl> said:
>
>> For documents meant for reading from displays font expansion makes no
>> sense anyway (actually, in print it also is somewhat debatable as
>> abusing it makes things worse).
>
> Not sure I agree with you here (or I at least don't understand you).
> Font expansion creates more readable paragraphs by producing lines with
> smaller \badness -- surely that's an advantage even on screen? Where
> does the nonsense come from? (Distorted glyphs? Reading PDF on screen in
> the first place?)

Displacement (and scaling) are normally less that pixel differences. 
Differently scaled glyphs in successive lines might stand out more on 
screen. Maybe if we have > 600 dpi screens ...

Concerning the badness ... there is some wishful thinking involved I 
guess ... when hz was introduced in pdftex we did experiments at 
usergroup meetings (each participant got its own randomized copy so no 
shared experiences).

The outcomes were interesting:

- We hadn't told that it was hz and protruding, just wanted to know 
opinions about side by side comparison.

- HZ samples not significantly percieved better. We did statistical 
analysis and the conclusion was that the rating was random.

- Protrusion was seen as a bug.

- Texies came with remarks about wrong penalties and spacing being set, 
of limited hyphenation patterns being used, rivers being seen ... 
basically coming up with reasons why paragraphs next to each other 
looked different but not per se better.

- No differences were see in many cases (we used 600 dpi high quality 
prints).

- Extremes like stretched wide o's in one line and shrunken ones in the 
following ones went unnoticed (so much for claims to see details).

When Thanh and I discussed that (at a Dante meeting) with Hermann Zapf 
he made the remark that probably 99% of the readers would not notice the 
difference. He also suggested that stretching lines vertically in order 
to get rid of widow lines was probably more effective. (I must have the 
tests done afterward somewhere.)

What was striking is that the fact that one uses an advanced engine like 
tex does not mean that the user is capable of seeing the difference. If 
someone shows me two versions of a document probably I cannot say with 
100% certainty which one has used hz.

The fact that we have an engine that can do all these things does not 
mean that we should always use it (the same is true for font features) 
and it also does not mean that users (and readers) see it .. anyway, one 
can use expansions and have maybe better paragraphs but with the wrong 
font and a messy layout it does not help much.

So, it all depends on factors like the font, amount of expansion, 
quality of print, rest of the layout, widths of columns, language 
(length of words, hyphenation etc influences solution space).

ps. See tex/context/samples/zapf.tex.

ps. I tend not to buy books that use expansion and inter character 
spacing trickery to the extreme as it distracts me too much

ps. Keep in mind that with for instance protrusion (and to some extend 
expansion) changes the solution space so we don't get better output. 
Also, expansion happens on a per-line base.

Hans

-----------------------------------------------------------------
                                           Hans Hagen | PRAGMA ADE
               Ridderstraat 27 | 8061 GH Hasselt | The Netherlands
     tel: 038 477 53 69 | voip: 087 875 68 74 | www.pragma-ade.com
                                              | www.pragma-pod.nl
-----------------------------------------------------------------