- From: <bugzilla@wiggum.w3.org>
- Date: Thu, 15 Nov 2007 17:04:44 +0000
- To: public-sml@w3.org
- CC:
http://www.w3.org/Bugs/Public/show_bug.cgi?id=4788 ------- Comment #7 from johnarwe@us.ibm.com 2007-11-15 17:04 ------- Ginny is correct in comment 6 about the intent of this bug. Scanning the editors drafts for URI myself, I find more than 2 places to change. Since I am doing this, I am also generalizing the bug to fully align with RFC 3986 terminology; anyone disliking that can ignore the extra changes listed here or move them into a separate bug. 4.1 Packaging, the example (2 places) from: "...smlerr:localizationid=xs:anyURI URI identifying the translation resource..." to: "...smlerr:localizationid=xs:anyURI URI reference identifying the translation resource..." 4.1 Packaging, the example (2 places) from: "...<alias> * xs:anyURI An URI by which..." to: "...<alias> * xs:anyURI An absolute URI by which..." 4.1 Packaging from: "...Each of these URIs serves as a name ..." to: "...Each of these absolute URIs serves as a name ..." 4.1 Packaging from: "...Typically it is a URI , an XLink..." to: "...Typically it is a URI reference, an XLink..." 4.2 Interdocument References from: "...If the URI in such a reference is equivalent ..." to: "...If the URI reference in such an interdocument reference is equivalent ..." (I realize this collides with the much-anticipated idr proposal, can be moved into one of those bugs so it is not changed twice w/o objection from me) 4.2 Interdocument References (2 places) from: "...is equivalent to the URI in an alias..." to: "...is equivalent to the absolute URI in an alias..." (ditto, possible collision) - several more spots where naked "URI" is used around aliases elided 4.2 Interdocument References from: "...written as a relative reference, URI coming..." to: "...written as a relative reference, the URI reference coming..." 4.2 Interdocument References from: "...The other URIs in the example..." to: "...The other URI references in the example..." 5.3.1 URI equivalence from: "...To determine whether two URIs are equivalent..." to: "...To determine whether two URI references are equivalent..." 5.3.1 URI equivalence from: "...corresponding characters in the URIs . ..." to: "...corresponding characters in the URI references. ..." 5.3.1 URI equivalence from: "...relative URI is tested for equivalence with another URI..." to: "...relative reference is tested for equivalence with another URI..." Note that here we actually have several specification problems. I suspect we want them in a separate bug, but I will list them here to get feedback on whether or not to open a new bug. - This text talks about comparing relative refs. RFC3986 6.1 says relative refs should not be directly compared; they should be converted to their target URIs and the target URIs should be compared. - Relative refs have an optional fragment component. When we are doing alias checking and use URI equivalence, there are two ways to look at it, but neither are discussed directly: 1. We first find the document, by comparing target URIs _after removing_ any fragment component, and then separately compare the fragment components if any of the document URIs are equivalent. 2. For any given ref, we generate a series of target URIs by taking each document alias in turn and appending the sml ref's fragment component to it, and compare each of those to the relative ref's target URI. Essentially, read strictly as currently written, an alias (always absolute URI which implies NO fragment component) would NEVER match any URI ref containing a fragment component. 5.3.5 is also contaminated. 5.3.3 Document aliases from: "...Each alias contains a URI. ..." to: "...Each alias contains an absolute URI. ..." 5.3.4 Relative references from: "...set is a relative URI , then the [base URI]..." to: "...set is a relative reference, then the [base URI]..." 5.3.5 Resolving Interdocument References from: "...If the URI representing..." to: "...If the URI reference representing..." 5.3.5 Resolving Interdocument References from: "...if the URI representing an interdocument reference ..." to: "...if the URI reference representing an interdocument reference ..." 5.3.5 Resolving Interdocument References (2 places) from: "...If the URI representing..." to: "...If the URI reference representing..." 5.4.1 URI prefix matching Whole section is ambiguous. As written, fragment components appear to be allowed and any fragment component would be part of the comparands, which I doubt is what we want. This could be spun off to a separate bug. Finally, this bug is targeted at SML only, but similar issues do exist in SMLIF (I just checked the editors' copy). Would folks prefer enumerating them here, or in a new separate bug?
Received on Thursday, 15 November 2007 17:04:54 UTC