- From: Jeni Tennison <jeni@jenitennison.com>
- Date: Fri, 5 Dec 2008 19:44:26 +0000
- To: "Dave Pawson" <dave.pawson@gmail.com>
- Cc: Toman_Vojtech@emc.com, public-xml-processing-model-comments@w3.org
Dave, On 5 Dec 2008, at 15:34, Dave Pawson wrote: > 2008/12/5 <Toman_Vojtech@emc.com>: >>> I also have one general (and heretical?) question about p:uuid and >>> p:hash: Why do we have these steps at all? > >> I will raise one argument against this idea myself: XProc is built on >> the principle of having a rich (and extensible) library of steps, not >> extension XPath functions. Furthermore, complex XPath extension >> functions are much more difficult to maintian than steps; the fewer >> of >> extension functions we have the better, IMHO. > > Putting it the other way round... > If the XSLT WG proposed adding xproc extensions within the XSLT WD, > wouldn't you be teed off? Not at all. Extensibility hooks such as that in XPath (which allows anyone to create extension functions) and XProc (which allows anyone to create their own extension steps) are all good, and should be used by anyone who wants to use them, WGs or not. We specifically want other WGs to define XProc steps that are useful for the kinds of processing they define (eg RDF-related processing) because they are experts in their fields and know the kinds of processing that they need to do. (Anyway, if we'd thought that creating extension functions was a bad idea, we'd never have had EXSLT, would we?) I think a better guide would be that any process that returns an atomic value (eg string, number) should be an XPath function; any process that returns XML should be an XProc step. Cheers, Jeni -- Jeni Tennison http://www.jenitennison.com
Received on Friday, 5 December 2008 19:45:01 UTC