Re: Page and line penalties

[David Feuer]

On 4/7/06, Joe Neeman <address@hidden> wrote:

> I think they have to be. In the page-turning case, if G is much smaller than
> F, the penalties for a bad turn get ignored. Conversely, if G is too big, we
> sacrifice nice spacing. I hadn't actually thought of this before, but it
> looks like this might be a main cause of the instability. I don't see a way
> around it while keeping penalties, though.

Comparing LilyPond to TeX, LilyPond faces a somewhat different
challenge.  Most of the time, TeX deals with lines that have many
words in them.  It can move a word from one line to another without
changing things too much.  It can even hyphenate words to make the
change even smaller.  Its input is structured into paragraphs, which
give fairly frequent mandatory line breaks that are also good page
break points.  LilyPond doesn't have paragraphs to help it limit the
line breaking decisions, and it doesn't have nearly as much
flexibility for breaking lines.  Lines in LilyPond could be as short
as three or four measures, and breaking a measure lies somewhere
between extremely bad and unthinkable.  LilyPond will need to make
compromises.  I think one thing that could help would be to allow
users to give LilyPond an idea of how fast the music should be played
at different points.  The penalty for breaking between two notes
should probably depend on their real-life length in seconds.  A slower
piece will end up aesthetically better, while a faster one will bow
more to practical concerns.

David