Re: XSL FO keep-together=int Implementation?

However, I don't think this is portable, and the FO processor
might decide to start a new page anyway once it detects the block wont
fit completely onto the current page (as FOP 0.20.5 does).

I'm not sure what "portable" means here.

Are you asking for the XSL FO example that demonstrates the use case
or the algorithm for the use case?

[ooo]
> If I remember the discussion of this when XSL FO was being developed,
> the idea is that you have to iterate over the block sequence some number
> of times to really do this nicely.  A bad choice for break might cause
> really bad things to happen to keeps constraints, floats, and/or cause
> odd spaces
> on pages to obey the keeps.

Exactly. Unfortunately, the spec doesn't give hints what "really bad
things" should be avoided, in particular regarding "odd spaces".

> One way out of this is to keep track of "bad placement" of blocks.
> Beyond some threshold, you might re-place a subsequence of them to
> try to achieve a better flow of areas.

Yes, but this means either keeping a lot of data in memory or costly
backtracking.