What I tried to propose is mostly just that. Implementing some shortcut that at least treats all integer values differently from "always" with a constant penalty value. That gives FOP the opportunity to relax while still allowing the rather intuitive "always" not to relax thus providing both kinds of behaviours. Relaxing "always" might not be that intuitive/expected for some people. Introducing a configuration option makes the whole thing just more difficult. This way you can just tell the user to use some integer instead of always if he wants the keeps to be relaxable.
Treating integer values differently from "always" is the half way to the full implementation but still can be implemented with reasonable effort without compromising any future improvements. The only problem I see here is that DocBook may not allow that kind of control over the keep values. For everyone else, that's just a quick search and replace in the stylesheet. On 13.07.2006 15:13:59 Chris Bowditch wrote: > Jeremias Maerki wrote: > > > Fabio Gianetti made a good comment [1]. I answered like this [2]. I'm > > currently thinking about how best to implement this. To keep it simple > > for the moment, we could implement "always" like before but > > remove/disable the overflow recovery I've implemented. That way, the > > content would again overflow. All integer values could be implemented as > > penalty=999 for the moment (thus allow some relaxing), at least until we > > have a good scheme about mapping integer keeps to penalty values like we > > started to discuss some time ago. However, this would disable the > > possibility to shove an element ahead n pages in the hope that there > > will be a page that the element fits on (the purpose of the overflow > > recovery). But that will be a very rare thing anyway, so I don't think > > there's any harm. Any objections? > > Well until integer values for keeps are implemented I object to > implementing always such that it generates overflow on a page. The > ability to error or clip gives little comfort to users either. > > Personally I prefer the Renderer to relax the keep-together="always" if > required. Fabio seems to suggest in his post that it is implementation > dependent what happens in this situation. Maybe we could have a > configuration option? > > <snip/> > > Chris Jeremias Maerki
