- From: Anne van Kesteren <annevk@annevk.nl>
- Date: Mon, 15 Oct 2012 11:08:26 +0200
- To: Darrel Miller <darrel@tavis.ca>
- Cc: www-archive@w3.org
On Sun, Oct 14, 2012 at 12:08 AM, Darrel Miller <darrel@tavis.ca> wrote: > When you say "everywhere", are you also including URIs used in media types > like text/uri-list, collection+json, hal+json, RDF/XML, VoiceXML ? Or are > you just interested in defining the behaviour of web browsers? There was this case on the HTTP mailing list a while back: Location: /hello world 1. It's invalid because Location does not allow relative references. 2. It's invalid because relative references cannot contain a space. Non-browsers also handle this case however (test it in curl for instance), because that is how these things work. They leak. > I'm really not sure how it is reasonable to specify how every possible > user-agent should "process" an invalid URI. 1. Many user agents already do. 2. It seems unreasonable to me to have distinct processing libraries for what is essentially the same data type. If I wrote an RDF crawler it sure would be nice if the code for the Location header used the same code path as used for URLs in say the RDF data. And again, these things leak. If you have a RDFa processor on top of the browsers' DOM it would suddenly handle more URLs than the one based on the server? That seems unsatisfactory to me and contrary to the value we get out of standards. -- http://annevankesteren.nl/
Received on Monday, 15 October 2012 09:08:54 UTC