Differing region-body ipd values, esp. for table column width

Dear XSL Editors:

I have two related sets of questions, one more general, the other more

As far as I can tell, the XSL-FO 1.1 WD allows pages within the same
page-sequence to have differing inline-progression-dimension values for
region-body (and therefore main-reference-area).

1. Is this the intent of the specification? I assume the answer is yes, but
wanted to clarify it because: a) This capability is difficult to implement
:-) b) I can't think of any use cases for this capability, c) Since XSL-FO
is derived largely from CSS, and this is an issue that doesn't affect CSS,
there seems a slim possibility that it is a mere oversight, and 4) (most
importantly) it raises the specific issue addressed below.

2. Would an implementation that does not allow this capability be considered
a conforming implementation? I assume that the answer here is "no". The
capability could be disabled either by a) throwing an error when this
condition occurs, or b) using the lowest-common-denominator approach of
using the smallest region-body ipd for pages in a page-sequence.

It is tempting to think that table column widths should be consistent from
one page to the next. However, if my assumptions about the general questions
above are correct, then they would not necessarily be so, especially where
percentages or proportional-column-width() values are used for column-width.
I don't think any great confusion arises from this for tables with fixed
layout. However, once the door is open for column-widths to differ from one
page to the next, tables with table-layout="auto" might also be affected,
*even when the region-body ipd does not differ from one page to the next*.
The 1.1 WD states, for this property "Use any automatic table layout
algorithm". However, since this was lifted straight out of CSS, the meaning
is unclear to me. The word "any" by itself would seem to allow differing
column-widths from page to page, even if all region-bodys have the same ipd.
The confusion arises from whether such an algorithm would qualify as an
"automatic table layout algorithm", or whether intrinsic to that definition
is the idea that column-widths should be the same from one page to the next.
A conforming CSS application could not (I think) vary the column widths at
all within the same table.

Thank you in advance for any guidance you can provide.

Victor Mote

Received on Monday, 25 April 2005 18:40:23 UTC