- From: Henri Sivonen <hsivonen@iki.fi>
- Date: Sat, 24 Jan 2009 15:45:19 +0200
- To: Robin Berjon <robin@berjon.com>
- Cc: Larry Masinter <masinter@adobe.com>, HTML WG <public-html@w3.org>
On Jan 24, 2009, at 12:53, Robin Berjon wrote: > On Jan 23, 2009, at 23:23 , Larry Masinter wrote: >> Secondly, "about:" URIs *do* have a clear retrieval path, and there >> is some expectation that, should you type one into a browser, you >> would get something meaningful; in this case, there is no >> requirement (or it would be unreasonable to expect) a browser to >> present something meaningful should one accidentally type >> about:sgml-compat into the address bar. > > Nothing prevents us from specifying what it should return (I'm > guessing the empty string, or a well-formed external subset with no > information items depending on how we'd like to tackle that). In text/html parsing as specified by HTML 5, the system identifier is not a URI. It is just a string that can be compared for equality with strings hard-coded into the parser. In application/xhtml+xml, the best way to achieve compatibility with XSLT and similar serializers is not to use a doctype at all. Thus, "about:sgml-compat" is *not* interpreted as a URI by any conforming HTML5 consumer. In my opinion, it is therefore unnecessary for it to be of the form of a URI in a registered scheme. The point of making it *look* like an absolute URI (i.e. have a colon in the magic string) is to avoid useless GET requests to URIs relative to the document URI in a situation where a piece of software goes and dereferences the magic string as if it were a URI. The point of suggesting "about" as the string before the colon was that due to pre-existing special use in browsers, it won't be feasible for anyone to register "about" as a URI scheme for another purpose. The 'tag' URI scheme is less suitable, because 'tag' URIs by their nature include non-mnemonic strings which make them harder to memorize. Furthermore, the date in the 'tag' URI scheme is dangerously close to being a version number, and one of the design goals was to avoid putting anything that resembles a version number into the doctype. The problem with 'urn' is that there are actual URN resolvers that map a subset of URNs onto dereferencable URIs. Even if the 'w3c' URN scheme went nowhere, finding out that it goes nowhere could still cause waste in theoretically possibly scenarios. Using about: addresses even that mostly theoretical case. -- Henri Sivonen hsivonen@iki.fi http://hsivonen.iki.fi/
Received on Saturday, 24 January 2009 13:46:02 UTC