- From: james anderson <James.Anderson@mecomnet.de>
- Date: Sat, 13 Feb 1999 22:05:20 +0100
- To: www-dom@w3.org
(in response to a note which appeared on xml-dev) After reading remarks to the effect, that all principal parser providers had released, or would soon release, parser versions with namespace support, I retrieved several versions to "see what I could see." four to be exact: xml4j, msxml/dcxml, sun, and oracle. I observed, in effect, four different language bindings for the (as yet unspecified) model for namespaces and decided, well, to go back to work. Upon seeing this message, however, I would like to note that it would be a welcome contribution from the respective implementers if they would offer some account as to why their respective interface is the way to go. In general, I was perplexed that the namespace URI appears to be on its way to becoming a property of the node (element or attribute), rather than of the node's name. What processing scenarios are people working eith for which this interface makes sense? : ? wholesale transplantation of "nodes" from one namespace to another (ie independent of the names)? ? matching nodes based on uri without regard to names? ? if one mutates a nodes namespace, does this affect the namespace of explicitly qualified attribute names? ? what sort of uri mutations are proposed and how is coherence maintained between the element name, attribute names (uri's and prefixes) where all are explicitly asserted. ? i didn't find facilities in all cases to operate on nodes with universal names, but where i did, i was uncertain how does one would use use something like getAttribute(uri x name) or getInheritedAttribute(uri x name) to find attributes in "no namespace"? I was also disappointed, that the interfaces, as best i could surmise, offer as yet not quite interoperable mechanisms for working with namespaces. (watch out, here comes another one of my tables; please set your windows on "wide" :) msxml/dcxml (betatwo) sun (ea2) oracle (1_0_0_3) ibm* (2.0.0) encoded name: Node.getNodeName, Node.getNodeName NSAttribute,NSElement.getQualifiedName Namespace.getName local part: IXMLDomNode.getBaseName, NamespaceScoped.getLocalName NSAttribute,NSElement.getLocalName Namespace.getNSLocalName namespace URI: IXMLDomNode.getNamespace, NamespaceScoped.getNamespace NSAttribute,NSElement.getGetNamespace Namespace.getNSName prefix: IXMLDomNode.getPrefix, ? ? ? universal name: ? ? NSAttribute,NSElement.getExpandedName Namespace.getUniversalName expanded name: ? ? ? Namespace.createExpandedName (* with whom i commiserate for having taken appendix a seriously :) Looks like the DOM comittee is in for an interesting week... How about an addenda to dom-level-1 which is just the java languange binding for the (as yet unspecified) namespace-aware name/attribute/element interface, also to include some arguments for the decision to model names within the application either as atomic or as explicit (string X string).... ? Chris Lovett wrote: > > An interesting thread. First, the DOM committee is addressing this issue > this week. IMHO the degree in which XML namespaces succeed will determine > the breadth and depth of the success of XML in general, and not just XSL - > so I eagerly await what the DOM committee comes up with. > > We have a namespace implementation in our DOM which we are shipping in IE5, > and I think IBM and Sun also have a solution. I didn't fully understand all > the arguments presented here - but our experience is that although > namespaces are not trivial to implement in an efficient manner, it is > doable. >
Received on Saturday, 13 February 1999 15:59:24 UTC