From: Colin P. A. <co...@co...> - 2005-02-25 14:26:35
|
>>>>> "Eric" == Eric Bezault <er...@go...> writes: Eric> I was about to reply to your previous message. I still haven't seen it. There definitely appears to be a problem with the mailing list at the moment. Eric> for this matter it is probably better to wait for Franck's Eric> approval than mine. I think so too. Eric> From what I understand though, having it in Eric> library/xml/xpath/xpointer shows the dependence between Eric> xpointer and xpath. Actually, that's only an implementation dependence (although some XPointer schemes (such as the unapproved xpointer() scheme, and the gexslt:xpath() scheme which i devized) DO depend upon XPath, the framework, and the current approved schemes (element() and xmlns()) don't depend upon XPath). Eric> I were to use the XML library and wanted to see if it Eric> supports XPointer without looking at the doc I would most Eric> likely try to see if there is a cluster Eric> library/xml/xpointer. My point was, if we wanted to have an implementation for the tree cluster (XM_NODE etc.), then it might be better to have an interface directory (xml/xpointer) plus implementation directories (xml/xpath/xpointer, xml/tree/xpointer). And even if we cannot come up with a common interface (but I think that might be possible, using generics), then the same cluster structure might still be appropriate, with xml/xpointer just containing a README.txt file, mentioning where the implementations were). But I don't know how feasible an implementation based upon xml/tree will be, as i am not at all familiar with it. Looking at my class XM_XPATH_XPOINTER (and if we decide to move the cluster up a level, abandoning the idea of multiple implementations, then I would rename this to XM_XPOINTER), I suspect that only two generics (possibly a third, constrained, generic might be necessary) would be necessary for a common interface. -- Colin Paul Adams Preston Lancashire |