I'm currently working on overconstrained geometry (5.3.4, XSL 1.0) and I'm unsure what to do with the following situation:
Assume a reference IPD of 5in. Assume an fo:block with start-indent="3in" end-indent="2.5in". --> 5in - 3in - 2.5in = -0.5in (for content area IPD) inline-progression-dimension is defined to treat negative lengths as "0pt" in 7.14.5. I can see that this applies to the property and not (directly) to the trait. Still negative values for the i-p-d trait seem to be unwanted. 5.3.4 dictates that in this case that start-indent + end-indent + content IPD = reference IPD. Now, if the content IPD is forced to "0pt", the above equation doesn't fit anymore and 5.3.4 says that end-indent is adjusted accordingly. In this case, end-indent is set to "2in". So far, everything is clear IMO. This was just to get you started. Now, considering a content IPD of "0pt". What shall we do with it? I'm currently letting it through. This produces block areas with very large BPD due to the many breaks, i.e. each break is chosen as concrete break. Text is painted outside the content rectangle. I know such a situation is almost certainly a bug in the stylesheet and this situation somewhat esotheric but still... I'm not sure what to do. The other idea is to skip the FO if the available IPD for its contents is 0pt. If anyone has some cool ideas, I'd be grateful, but don't waste too much time thinking about it. :-) Jeremias Maerki
