- From: <Toman_Vojtech@emc.com>
- Date: Fri, 22 Aug 2008 10:27:00 -0400
- To: <public-xml-processing-model-wg@w3.org>
> > Imagine that library.xpl contains: > > <p:library ...> > <p:declare-step name="xinclude-and-xslt">...</p:declare-step> > </p:library> > > Consider the following imaginary command line syntax: > > $ calabash -isource=- -oresult=- -s xinclude-and-xslt library.xpl > > This tells the processor to load "library.xpl" and run the step named > "xinclude-and-xslt". All straightforward. > > Now suppose I want to just run the "add-attribute" step. I could wrap > it in a pipeline, put it in a library, give it a name, and run it. But > since the library for the XProc builtin steps is always loaded (at > least conceptually), why do I have to do that? Why can't I just say: > > $ calabash -isource=- -oresult=- -s add-attribute \ > match=div attribute-name=foo attribute-value=bar > > I can't because the standard p:add-attribute step doesn't have a name. > Can't you do something like this on the command line: $ calabash -isource=- -oresult=- -s p:add-attribute \ -xmlns:p=http://www.w3.org/ns/xproc \ match=div attribute-name=foo attribute-value=bar You don't need to give the step a name in this case, and it is also more consistent with the way you would invoke the step in an XProc pipeline. The step names are "local" in nature - if you import a library in a pipeline, the step names are invisible in the target context, so the pipeline does not see the names of the imported steps, only their types. So, in my opinion, I think that using step types instead of names on the command line is more correct. Regards, Vojtech
Received on Friday, 22 August 2008 14:27:55 UTC