Sjoerd Visscher wrote: > > Sam Ruby wrote: > >> Apparently, consuming tools are welcome to aggressively substitute >> references to the enclosing parent document of any element for any >> references that, when resolved according to xml:base, differ from that >> xml:base only in ways that deal with normalization and fragment >> identifiers. This can only cause confusion if the xml:base in effect >> differs from original xml:base of the document (i.e., the URI used to >> retrieve the document in the first place) in ways other than the >> fragment identifier. > > You've nailed it.
Cool. Took me long enough. >> Note that I'm sidestepping all questions about who is right or wrong. > > I totally agree, there is no right or wrong here. The established usage > of a base URI is so different from what Roy is saying that he shouldn't > have changed the RFC in such a way. (The RFC is even an Internet > Standard, defined as "a specification that is stable and well-understood.") I'm sure that we can find established usages where the consuming tool has various degrees of "agressiveness". Pure speculation on my part, but now that I have been able to see how limited in scope this issue is, I can see where spec authors might have had trouble outlawing one behavior or another, and decided that publishing a spec within a reasonable timeframe was of greater value than resolving this issue. >> The recommendations produced by the feed validator will focus on the >> areas where the user is most likely to stumble into this problem. It >> seems to me that the largest problem area is at the feed level, and the >> recommendation will be to either make xml:base at the feed level match >> the URI from which the feed was loaded, or (paradoxically enough) to >> reference a resource that you are unlikely to directly reference later >> in the document. Referencing a parent directory of any given document >> is OK, what's important is that it isn't the document itself. > > Yes, although I wonder how you would test for "unlikely to directly > reference later". I will only test for actual references that meet the criteria described at the top of this note. When I encounter such a condition, I will emit a warning containing a one line message accompanied by a link. That link will take the user to a web page that repeats the one line message, provides an explation of that message (probably close to what I said at the top of the page), provide a recommendation (probably close to what I said in the second paragraph above this one), and a link to the feedvalidator mailing list. - Sam Ruby
