debates about the "paragraph-at-once" algorithms (which have
never worked in a satisfying way to me when I have needed to
typeset with TeX).
Accepting that no algorithm will produce perfect results, the question
isn't whether paragraph-at-once algorithms obviate the need for human
intervention, but whether they overall produce better results than
setting every line individually without regard for the surrounding
paragraph. That is, which system leaves the human with less work to
do?
Being a long-time TeX user well before I seriously took up groff for my
work, I must confess that the paragraph-at-once algorithm isn't bad at
all once you understand how to tweak its output. Fixing unpleasant line
breaking points usually involves inserting just one additional
hyphenation point, or occasionally preventing an automatically executed
hyphenation. After a little training, these points are easily
identifiable. So, it's not a question wether the algorithm [alone]
produces perfect results, it is very much a question of the quality of
the hyphenation patterns for a given language upen which the algorithm
is executed. In addition, the situation is not the same for every
language. Some languages are very productive in terms of long compounds,
e.g. German. Take this:
"Rindfleischetikettierungsüberwachungsaufgabenübertragungsgesetz",
https://en.wikipedia.org/wiki/Rinderkennzeichnungs-_und_Rindfleischetikettierungs%C3%BCberwachungsaufgaben%C3%BCbertragungsgesetz
This is possible because of highly productive German word formation
rules. In contrast, Altaic languages, being agglutinative in nature,
like Mongolian and Turkish, can produce quite long words due to the
plethora of grammatical suffixes that can be attached, even recursively.
See https://en.wikipedia.org/wiki/Longest_word_in_Turkish.
Best regards,
Oliver.
--
Dr. Oliver Corff
Mail:oliver.corff@email.de