- From: <keshlam@us.ibm.com>
- Date: Tue, 6 Jun 2000 12:31:28 -0400
- To: Paul Grosso <pgrosso@arbortext.com>
- cc: xml-uri@w3.org
>Whoa, you are talking about schema-validity with assumptions that namespace >names point to schemas, etc. Nope. I'm talking about namespace-aware validation, which schemas will support, combined with the Absolutize proposal for namespaces. Consider what happens if the shema's declaration for the namespace, and/or the instance's declaration for the namespace, is expressed as the same relative reference. In that case, if the two have the same base address at the point of declaration they match and the document is valid; if they don't, they don't match and the document doesn't validate. If an XBase is specified, then the base address at that point may differ depending on whether or not XBase is supported by the processor which read that document. In the above example, where the declaration uses the same string, adding XBase support changes the result of the absolutize operation. This could cause a document which was validating to no longer validate. You can also reverse the scenario to create documents which validate when XBase is available but fail when it isn't. Maybe the right answer really is "A relative reference refers to a family of resources; if you care which member of the family you get, you had no business using relative references in the first place." But personally, I don't like designing tools that both cost more to operate and make injuring yourself easier unless there's a Pressing Need to do so. Especially when I think about how much time I'm going to have to spend explaining to folks that if they point the gun at their foot and pull the trigger, they can't complain to customer support. ______________________________________ Joe Kesselman / IBM Research
Received on Tuesday, 6 June 2000 12:31:50 UTC