Re: too small inter-word spacing

[Joerg van den Hoff]

On Sep 08 2008 (Mon, 14:03), Martti Haukijärvi wrote:
> 2008/9/5 Joerg van den Hoff <address@hidden>:
> > recent documents: do you not see incidences of way too tight spacing 
> > between words? do you feel the spacing is
> > on average a) optimal, b) to loose, c) to tight?
> 
> Most of spacing is OK, but often (a couple of times in every page)
> spacing is too tight. This is visible especially at the end of
> sentences, dots being too close to next word. Perhaps this is problem
> with those languages which have a lot of long words, like Finnish? In
> English words are shorter and hyphenation seems not to be used so
> routinely as in Finnish, for example.  Perhaps the hyphenation penalty
> should be lowered for such languages.
> 
> -- 
> Martti Haukijärvi

martii,

thanks for responding. my assessment is similar to yours: spacing OK
most of the time. but even here I think the spaces are frequently
rather tight, i.e. generally compressed below the "optimal" standard
width of a space. On the other hand, I seem to find _no_ obvious
instances of significantly expanded spaces at all.

I too, see many instances of too small inter-sentence space (as a
special case of the inter-word problem). it seems the same with other
punctuation, notably commas: the space after the comma is frequently way
too small. here I think the problem of too much compression becomes
especially apparent, but the problem is there otherwise, too.

so I still believe that there is some room for improving the current
settings controling the paragraph breaking behaviour.

my guess concerning frequency of occurence is that it has to do with the
"graininess" of the text in terms of how many words end up on the same
line. if this number becomes small the "mis-behaving" becomes apparent.
so it should happen with many long words and/or with large point sizes
(this is where I see it).

jeff provided some helpful remarks concerning where the relevant
settings are found (in z14.c). my playing around with them up to now
showed that increasing "MAX_SHRINK" from its default 4 to 32 and then
recompiling does for some reason or other improve the spacing (and leads
to significantly more but not excessive hyphenation). this effect was
unexpected for me (in view of jeff's explanation of what that parameter
probably should do (give the factor of max. compression of a white
space) as well as simply because of its name...), but that's how it is.

of course I believe that this fiddling should not really be necessary in
order to "get it right". rather it should be handled silently by the
formatter. I assume that everybody agrees that the 'optimal' spacing is
the one defined by the inherent width of the space character. now, if
my subjective observation is right, that lout currently quite generally
only compresses spaces but seemingly never (or only _very_ rarely)
expands them and that rather drastic compression is nearly always
preferred to hyphenation/expansion, I would argue that this is not the
best possible compromise.

I understand that a rather complex algorithm is calculating the "best"
paragraph breaking and that there is no single "correct" solution and
that I'm arguing from a personal point of view. but anyway right now my
impression is that in effect lout more or less always compresses the
available spaces in order to avoid, both, hyphenation as well as
expansion.

which I would think is not right: too much compression definitely
compromises readability, too much hyphenation is a bit annoying, but
does no real harm, too much expansion is aesthetically unpleasant but
does no real harm, either. so I would argue for really limiting the
maximally possible compression and stop it much earlier (at the price of
finding some other solution via expansion and hyphenation) than
currently is the default.

what I've seen from lout up to now is a good design and ease of handling
many standard tasks superior to TeX and, of course, to troff. but both
these systems are _not_ infamous for bad typesetting. and looking at
their ouput clearly shows that they are consistently avoiding space
compression of a degree comparable to lout's. and if I pick at random
a few books from my shelf it's the same: looser spacing than lout's.


joerg