- From: Vasil Rangelov <boen.robot@gmail.com>
- Date: Mon, 10 Sep 2007 20:55:45 +0300
- To: <public-xml-processing-model-comments@w3.org>
>It doesn't fit conceptually with how XProc works (at the moment). ><p:document> isn't a step, so it doesn't have options: it's just a way >of referring to an external document from within a <p:input>. > >There's a possible extension of XProc some time in the future in which ><p:input> *does* hold nested steps. In that putative version, ><p:document> would be equivalent to <p:load>, and the href attribute on ><p:document> would be an attribute value template and thus able to take >dynamic values. But we don't have that facility now and working it all >out would be a major undertaking. Something for the next version, not >for this. OK. Point taken for p:load. But then what's the use of p:identity? The only use case I imagined was loading of documents within the pipeline without actually doing anything to them at that point, but that seems to be perfectly doable with p:load. What other use case is there in which p:identity is needed? >Speaking only for myself, I wouldn't say that I had a pressing need for >validate-on-parse or resolve-externals in other locations, given their >presence on <p:load>. Resolve-externals (or any equivalent to it) is not available anywhere. It should probably be at least added to p:load. >I can see the utility of p:validate-dtd to validate intermediate >documents, but every language I can think of that is primarily a >DTD-driven language has a RELAX NG equivalent. And if it doesn't >already, it's easy enough to create one. XProc doesn't seem to provide an easy way of creating RNG or XSDL files from DTD. And the pipeline author may not be the language's author (i.e. the creator of the DTD), so such sort of utility may be useful in those cases. Also, if such a step is not provided by the standard, something tells me it's going to be (in)consistently implemented across XProc processors anyway, since all XML parsers support DTD, right? I mean, if that's not the case, the validate option on p:load wouldn't have been required for conformance. Regards, Vasil Rangelov
Received on Monday, 10 September 2007 17:56:20 UTC