Talk:HttpRange14Requirements

JR: David Booth added the following comment, which JAR didn't understand. JAR moved it to the discussion page to avoid interrupting the flow of the main page.

DB: At present this requirement is stated more strongly than it should be. As stated, it would prohibit ambiguity on the "document versus subject of the document" axis. But it is not necessary to prohibit that particular kind of ambiguity universally, just as it is not necessary to prohibit any other particular kind of ambiguity universally. Rather, the necessary requirement is that it must be *possible* for a publisher to avoid this ambiguity (or any other kind of ambiguity) for a target class of applications. For example, when a publisher wishes to support consuming applications that need to distinguish a document from its primary subject, then there must be a way to do that. More specifically, the publisher must be able to publish data that enables its target class of consuming applications, without special knowledge, to distinguish between the document and the document's subject (or between two cases on any other axis of potential ambiguity along which the publisher has chosen to differentiate). -- David Booth 14:42, 25 October 2011 (UTC)

JR: After reading this JAR changed "if the entity described by the document/IR is the document/IR" to "if the URI admits a common interpretation in both cases" hoping to sidestep whatever issue it was that David was trying to raise.

JR: On thinking about this some more and I can't find any flaw in what I said before. If a URI refers to the document/IR at the URI (true by assumption A0) and the document/IR at the URI describes what the URI refers to (true by assumption B0) then the document/IR is self-describing. This seems unavoidable. So there is not even a need to talk about "the entity" because we know what it is. There is nothing in this logic that rules out someone believing that an earthquake is a document, and a self-describing one in particular, so the "punning" / "shadow properties" / "ambiguity" / "overloading" style approaches are supported.

In a situation where the agreement is only that the URI refers to either the document/IR at the URI or an earthquake, desideratum A0 is simply not met, since the agreement gives no way to conclude that it refers to the document/IR.

JR: New wording "If a URI is one that can result from either process A or B, then it must be possible to determine which one applies (or if both do, then determine either that one applies or that the other applies)." -- it is hard to word this accurately, and even harder to explain the wording, but this is getting closer to the mark I think.