- From: Jonathan Rees <jar@creativecommons.org>
- Date: Tue, 9 Feb 2010 08:58:20 -0500
- To: noah_mendelsohn@us.ibm.com
- Cc: Dan Connolly <connolly@w3.org>, Ben Adida <ben@adida.net>, www-tag@w3.org
There are two different cases, and we probably ought to consider them separately. 1. In my original message the same # URI is used to refer to two different things: An element (as the text/html media type reg would have it) and something else, say a person (as RDF would have it). 2. In the situation you describe the # URI is only used to refer to the non-element; there is no element with the given name. Your link checker would detect a problem in case 2 but not in case 1. I'm not sure what business a link checker has in looking at URIs that are only used in non-hypertext contexts (e.g. rel= but never href=), but I guess that would be its prerogative. In case 1 the risk would only be to an agent that knew about the RFC, and from the existence of an HTML representation, inferred that the URI referred to an element; and then later encountered RDF that put the URI's referent in a class that's disjoint with element (e.g. foaf:Agent), thus deriving a contradiction. FOAF (http://xmlns.com/foaf/0.1/) follows pattern 2 (which is unfortunate since on following one of its # URIs you're taken to the top of the file, not to the specification of that # URI). I have not found an instance of pattern 1 yet, but would be surprised if there weren't one. (By the way I just read the fine print in the 'best practices' note http://www.w3.org/TR/swbp-vocab-pub/ and it advises using 303 for namespace documents in the situation where HTML/RDF conneg is being done. E.g. Void (http://rdfs.org/ns/void) follows this pattern. The document at the end of the 303 is itself at risk, but the chance that in this case there are non-element # URIs based on the 303 target URI seems vanishingly small.) I will take the answer to my original question as "no one who reads www-tag has thought about updating the media type registration". Perhaps it's not so important, as in five or ten minutes of searching I found fewer instances of HTML/RDF conneg than I expected to, and future instances may adhere to the SWBP note. Jonathan On Mon, Feb 8, 2010 at 7:07 PM, <noah_mendelsohn@us.ibm.com> wrote: > Dan Connolly writes: > >> [Noah Mendelsohn writes:] >> > As to amending the media type specification: in principle I might be >> > concerned, precisely because people could have invested in code that >> > interpreted the failure to resolve as an error (at least in the same >> > spirit that 404 is an error). >> >> What failure to resolve? Could you elaborate the case you have >> in mind? > > I'll be glad to try. I can at least propose something hypothetical that > might be suggestive of the concern. Imagine that someone had built an > agent that helps debug broken links, perhaps for use in conjunction with a > content management system. Per the applicable specs, if > http://example.com/xyz.html#frag returns text/html, then the URI is a > reference (or attempted reference) to an element in the file, and the tool > should include it in a list of broken links if it doesn't resolve. If > instead the resource owner uses this as a reference to some other > secondary resource [2], it will get incorrectly counted as a broken link. > > I think this is also in the spirit of the concerns that Eric Bowman > raised. I don't know if you'll find this particular example convincing, > but it's the sort of thing I have in mind. > > Noah > > [1] http://lists.w3.org/Archives/Public/www-tag/2010Feb/0058.html > [2] http://www.w3.org/TR/webarch/#def-secondary-resource > [3] > > -------------------------------------- > Noah Mendelsohn > IBM Corporation > One Rogers Street > Cambridge, MA 02142 > 1-617-693-4036 > -------------------------------------- > > > > > > > > > Dan Connolly <connolly@w3.org> > 02/05/2010 06:30 PM > > To: noah_mendelsohn@us.ibm.com > cc: Ben Adida <ben@adida.net>, Jonathan Rees > <jar@creativecommons.org>, www-tag@w3.org > Subject: Re: HTML media type vs. # URIs that do not > identify document elements > > > On Fri, 2010-02-05 at 16:54 -0500, noah_mendelsohn@us.ibm.com wrote: >> Dan Connolly wrote: >> >> > You say that like it's a bad thing. >> > >> > i.e. what's "wrong" about that? >> >> > [..] >> >> > Why would browsers do anything different >> > from what they do now? >> >> Perhaps I wasn't clear: I have no problem at all with what the browsers > >> are doing. >> >> I believe Jonathan pointed out a use case in which the semantic Web >> community was serving text/html documents, with fragids used for > purposes >> that were not in conformance with the applicable media type > specification. >> You acknowledge that's the issue, where you say: >> >> > I wrote about this in a 2006 workshop paper... >> > >> > [[ >> > In order for this to work with documents published both in RDF/XML and >> > XHTML, the XHTML media type specifications may need to be ammended so >> > that authors can opt out of the section-of-the-document meaning of >> > fragment identifiers that they publish. For example, the profile >> > attribute from section 7.4.4.3 Meta data profiles of the HTML 4 >> > specification[HTML4] seems like a reasonable opt-out signal. >> > ]] >> > -- section Fragments as sections vs. people >> > http://www.w3.org/2006/04/irw65/urisym#docdata >> >> Right, but there's at least some damage in the meantime, with content > out >> on the Web that's in violation of current applicable specifications. I'm > >> not claiming the Web will crumble tomorrow over this, but I don't think >> it's a good thing. I used the browser example merely to point out the >> kind of damage that might, at least in principle, be observed. > > But what you call "damage" looks like stuff working as expected > and as designed, to me. I don't see the point at all. > > >> As to amending the media type specification: in principle I might be >> concerned, precisely because people could have invested in code that >> interpreted the failure to resolve as an error (at least in the same >> spirit that 404 is an error). > > What failure to resolve? Could you elaborate the case you have > in mind? In the case that Jonathan brought up, there's no > failure to resolve. > > >> In practice, it's hard for me to imagine >> that there would be significant trouble for anyone, and something like a > >> profile attribute seems like a reasonable way to signal the opt-out. >> >> Noah > > > > -- > Dan Connolly, W3C http://www.w3.org/People/Connolly/ > gpg D3C2 887B 0F92 6005 C541 0875 0F91 96DE 6E52 C29E > > > >
Received on Tuesday, 9 February 2010 14:06:50 UTC