On Wednesday 10 July 2002 06:25, Joseph Kesselman wrote: > Which is why it's being corrected in XSLT 2.0.
Yes, that is a Good Thing. Unfortunately, we're stuck with 1.0 for another year or so. (BTW, has Xalan made any moves towards supporting 2.0? Will that be Xalan3 or something?) > But a "quirks" solution would be no better than the extensions we have > now. Less so, if anything, because it hides the details of where you're > departing from the standard and thus makes porting the stylesheet to > another processor with different quirks much, much harder. Only people who know that they "want" quirks would enable it. Hopefully this would be an "enable at your own risk" thing. > "Quirks" simplifies things only as long as you never change processors, or > change only to processors that have _IDENTICAL_ quirks. If the quirks don't > match *exactly*, it rapidly becomes a maintainance disaster. This particular quirk (casting RTF->node-set) *is* the same with other processors. That's why I've requested it. Xalan is the only one that doesn't do it. The best way I can think to do it would be to add a namespaced attribute to the stylesheet root, like this: <xsl:stylesheet ... xalan:auto-rtf-to-nodeset="yes" xmlns:xalan="..."> To me, it's pretty obvious that a stylesheet is using extended functionality whenever I see a xalan:xxx or saxon:xxx name. Saxon and other processors all provide the ability to customize the processing using such attributes, for example to control the output escaping by a special attribute on <xsl:output/>. In general, I agree that random deviations from the spec are a Bad Thing, at least from a portability standpoint. But in this case, allowing deviation would make stylesheets *more* portable, not to mention easier to write. -- Peter Davis
