Cart
[Home] [By Thread] [By Date] [Recent Entries]
Jonathan Borden wrote: > It is not totally clear to me that the GI and DOCTYPE definition need define > the same 'type'. No indeed. Given the differences in the (multiple) roles of each, it requires tortured construction to say that what they define is in any sense at all the 'same'. Actually, the root element GI and the DOCTYPE name do not 'define' anything. Each is a lexical anchor or identifier to which content models, schemata, and processing strategies--which go further in attempting to define something--might be attached, which was my particular point. > That is to say, the GI of the root element, particularly when the element is > namespace qualified, points to the namespace as 'half' of the type and the > 'local name' which is qualified by the namespace, as the other 'half'. Taking 'namespace' here in the purely lexical sense employed by "Namespaces in XML", this is of course true, but it means no more than that, to a namespace-aware processor, the lexical anchor provided by the root element GI is simply one component of a larger lexical anchor. > DOCTYPE however, allows the document instance to have a particular association > of root element name and a DTD. Presumably the DTD _could_ change from > instance to instance. In practice, DTDs, when they do change, may change over > the lifespan of an instance, e.g. during edit mode, and are more properly > associated with the how the document is to be processed, rather than some > 'instrinsic' property of the instance itself. I suspect that you would agree > with this, as I am restating the 'local control' position. Yes, of course. To me the obvious way to accomplish this is by versioning expressed in the namespace portion of 'lexical' namespacing. > The question is whether a non-DTD hardwired replacement for DOCTYPE would be > useful http://lists.xml.org/archives/xml-dev/200201/msg01278.html Yes, of course, and RDDL is a mechanism well-suited for pointing to such DOCTYPEs and thereby asserting the association. You and I are clearly in agreement here, as I hope we continue to be when I insist that such a mechanism for associating DOCTYPEs or DOCTYPE replacements with both the lexical root element GI and with document instances exhibiting that GI needs to permit many-to-many (to-many) relationships. > The position that 'document type' has primacy over root element name is > articulated: > http://lists.xml.org/archives/xml-dev/200201/msg01518.html > (e.g. AF is an easy way to change the root element name). Again, I agree with what I understand you to say in the previously cited posting above, and by implication not with Steve in this posting, in that associations (again, let me stress, simple lexical associations) to DOCTYPE, or its replacements, are orthogonal to similar lexical associations to root element GI. It appears to me that you designed RDDL to deal well with such overlapping cases. > I posed the hypothetical question in order to help sort out this issue > ("Strategies for a lowly XML document"): > http://lists.xml.org/archives/xml-dev/200201/msg01355.html You propose one strategy, based on a PI, which may be useful. It is certainly not harmful, as anything which provides more information about an instance document to a potential processor of that document might well be helpful. As always, however, 'intent' expressed by a document creator is not binding upon potential processors of that document, who are free to ignore expressions of that intent, as PIs or otherwise. > Gavin feels strongly that the instance ought not dictate the processing: > http://lists.xml.org/archives/xml-dev/200201/msg01408.html He is right. > So perhaps the conclusion is that 'document type' is not a useful way to > direct processing. But often it is, given the particular purpose of that processing and the resources available to it which might not be known to the document creator. > The question remains whether 'instance centric semantics' has meaning beyond > 'root element name' but this remains a theoretical issue -- that is to say, > the user of the document 'knows' what he or she wants to do with it, and has > final say. I argue that there are no 'instance centric semantics', only instance syntax. The semantics are elaborated through the processing of a particular instance on a particular occasion by a particular operation. > When the author of a document has a specific and single intended 'semantics' > to be associated with the instance, the markup ought be highly constrained and > reflex this intent. In such cases the root element name does need to know, or > have the means to find out about, all its offspring. It is in such cases where > XML namespace qualification is particularly helpful, because the > root element namespace name/URI is a means of obtaining information about the > namespace and hence the intended semantics of the root element. OK, but I argue that this is not the general case. This is an instance of specific agreement between document creator and document processor utterly outside of the content of the instance document. The namespace mechanism you describe here is simply the lexical signal to both the document creator and document processor that this instance document is subject to their a priori agreement. Respectfully, Walter Perry
|
