- From: Peter B. West <lists@pbw.id.au>
- Date: Wed, 31 May 2006 08:01:05 +0100
- To: xsl-editors <xsl-editors@w3.org>
> > The XSL FO SG discussed this issue again at our recent f2f. > > We reconfirmed our position on this issue which is that there > will be no change to the 1.1 spec in this area. > > We understand that there are two ways inheritance could work > in the FO tree with respect to the properties in the layout-master-set. > Each of the two ways is necessary to support certain capabilities. > In scoping XSL 1.0, we decided to restrict ourselves to a single > simple inheritance model that always follows the structure of the > formatting object tree (section 5.1.4); that limitation continues > in XSL 1.1. We have always had plans to allow the other way as part > of a larger effort to support layout-driven composition requirements, > but we determined that work to be out of scope for 1.0. The reasons > we picked the inheritance model we did instead of the other one are: > (1) we had to choose one for 1.0, and most of the time one wants > inheritance from the content, and (2) though out of scope for 1.0, > we knew we would want to do layout-driven formatting at some point, > and this would require much more capability to "inherit" properties > in ways other than down the formatting object tree, so we decided > we'd leave all such capabilities for the future work. > My thanks to the editors for their valiant, if unsuccessful, attempts to enlighten me as to the meaning of "a single simple inheritance model that always follows the structure of the formatting object tree." Clearly I have been at cross purposes with the editors in arguing for what I understood to be "a single simple inheritance model that always follows the structure of the formatting object tree," because, all the while, I thought that the changes proposed would destroy such a model. Alas, I am still not entirely sure of the principles. I look at this familiar diagram - http://www.w3.org/TR/xsl/slice6.html#section-N10230-Pagination-Tree-Structure - and I cannot quite see how it works. As best I can determine, the regions of the simple-page-master, on the left of the great divide, inherit two properties and two properties only, viz. writing-mode and reference-orientation, from page-sequence on the right. This appears to be a form of cul-de-sac inheritance because, as near as I can determine from the editors' excellent exposition of the rationale, the simple-page-master and its regions create no further ripples on the inheritance of flows and static-contents, which inherit everything else in the usual uncontroversial manner. Pondering this mysterious effect, and considering another popular diagram from the Recommendation, to wit http://www.w3.org/TR/xsl/slice4.html#area-tree-sample I am beginning to suspect that there are undocumented quantum mechanical effects at work. I the above diagram, flows areas are pendant to areas from the troublesome regions. Now, as I have recently been reminded by the editors, area traits don't inherit. However, it would be extremely convenient if superpositions of the property "writing-mode" and the trait "writing-mode", and like-wise for "reference-orientation", were introduced. These superpositions could be collapsed into one state or the other as required from moment to moment during the reading of the Recommendation; a Heisentrait, so to speak, or Schrödinger's attribute. At the sub-atomic level, of course, this is all very simple, but it can be difficult to follow with the logic of macroscopic structures. It may surprise the editors to learn that many readers, and indeed implementors, lack even a rudimentary understanding of quantum mechanics. Witness my own misunderstandings of this change. At the very least the principles deserve a non-normative appendix. My congratulations on this ground-breaking work. Peter West -- Peter B. West <http://cv.pbw.id.au/> Folio <http://defoe.sourceforge.net/folio/>
Received on Wednesday, 31 May 2006 07:01:28 UTC