- From: David Brownell <db@argon.Eng.Sun.COM>
- Date: Mon, 14 Sep 1998 10:18:05 -0700
- To: xml-names-issues@w3.org
Section 6.1 on "The Insufficiency of the Traditional Namespace" seems rather broken. Since I've had occasion to study the "traditional" notions of naming in computer systems rather extensively, I was amused to be told there that they don't handle the problems being identified. It was clearly incorrect, and caused needless confusion. Names are "traditionally" composed of a context (sometimes implied) and a name. Two names (e.g. "High Street") are distinguished by context (e.g. "Bath" vs. "London"). Contexts are often named, and if so have contexts. ("Bath" in "Maine, USA" vs in "UK".) The assumption in 6.1 is that there is no such thing as a context in traditional naming within computer systems, which is very wrong. The notion of context is in fact quite fundamental, and is clearly present even in the first example there. Only by removing that notion (discarding the "attribute of author" or "element" contexts) can that example be made comprehensible. Even the XML spec doesn't support such usage, since the contexts for attribute and element names are clearly critical when validating. (The word "namespace" is used in several different ways, and usually means no more than which syntax is used to represent names in a given context. For example, integers vs words; or X.500 vs DNS. The latter two bundle syntaxes for context names, and can be combined in namespaces such as those allowed for email addresses.) I think that what "Section 6" wants to say is that while XML documents originally had two types of contexts (one for the element names, and for each element one context of attribute names) this is felt to be an insufficient data model in that (a) multiple element name contexts are desired, and (b) attribute name contexts shared between elements are desired This merits fixing, perhaps most expediently by deleting 6.1 and becoming clear on the two goals noted above. - Dave p.s. If desirable, I should be able to shake loose a good reference on "traditional" naming in computer science, which could be put into the references.
Received on Monday, 14 September 1998 13:20:51 UTC