- From: Martin J. Dürst <duerst@it.aoyama.ac.jp>
- Date: Tue, 25 Sep 2012 19:46:55 +0900
- To: Julian Reschke <julian.reschke@gmx.de>
- CC: Robin Berjon <robin@w3.org>, Larry Masinter <masinter@adobe.com>, W3C TAG <www-tag@w3.org>
On 2012/09/25 17:49, Julian Reschke wrote: > On 2012-09-25 10:31, Robin Berjon wrote: >> On 25/09/2012 04:29 , Larry Masinter wrote: >>> I think there was a group willing to consider the redefinition of >>> URLs in HTML5 as a local anomaly within HTML, in a way that didn’t >>> really affect any other format or application. >> >> My understanding is that Anne is working on an improved definition of >> URLs because he noticed demonstrable severe interoperability issues with >> tasks as deceivingly simple as parsing URLs. >> >> Has anyone in this thread taken if only five minutes to perhaps peruse >> the evidence and see if he might not have a point? I ask because I've >> given it a cursory look and what I've seen is ugly. > > Of course there is a point. The specs (RFCs 3986 and 3987) do not define > how to treat broken identifiers. Furthermore, references in HTML > definitively do require preprocessing (such as dropping leading > whitespace, or potentially rewriting query parts when not in UTF-8) > before they can be handled as URIs/IRIs. > > This is not a new discussion. Fully agree. Indeed, lots of attempts have been made to try and describe what browsers actually do with goop they find in a@href, img@src, and the like. If Anne can pull that off, then hats off to him. But given the current divergences between browsers, it may not exactly be easy. > I believe that this can be best handled by acknowledging that what HTML > uses are identifiers that need some level of sanitization before they > can be treated as URI/IRI (references). > > It appears that Anne's approach is to pretend that the RFCs are broken > and need to be completely replaced. This of course ignores that fact > that they are widely implemented outside browsers. > > What we IMHO need is a *precise* problem statement, and then a mapping > layer. I don't think the problem statement is too difficult. What Anne is after is implementation instructions for browsers. That's a good thing to have. But for somebody creating an URI or IRI, or creating an URI/IRI scheme, browser quirks can and should be irrelevant. It would be hopelessly confusing for them to look at Anne's document. Regards, Martin. > Also, it's not helpful that terminology from 3986 ("resolve") is used > for something else, leading even to more confusion. > >> ... > > Best regards, Julian > > >
Received on Tuesday, 25 September 2012 10:49:00 UTC