- From: Geoffrey Sneddon <foolistbar@googlemail.com>
- Date: Thu, 28 May 2009 15:39:53 +0100
- To: Anne van Kesteren <annevk@opera.com>
- Cc: "Sam Ruby" <rubys@intertwingly.net>, "Maciej Stachowiak" <mjs@apple.com>, "Roy T. Fielding" <fielding@gbiv.com>, "Larry Masinter" <masinter@adobe.com>, "HTML WG" <public-html@w3.org>
On 28 May 2009, at 14:57, Anne van Kesteren wrote: > On Thu, 28 May 2009 15:41:36 +0200, Sam Ruby > <rubys@intertwingly.net> wrote: >> I don't understand the "conformance with HTTP" part of the >> question. I >> believe that the current spec'ed behavior constitutes "a willful >> violation of the HTTP specification, which requires that the >> Content-Type headers be honored, despite implementation experience >> showing that this is not pratical in many cases." > > Currently they completely violate HTTP. By following the rules layed > out in HTML5 they could get much closer. (I agree that it is > probably better for this part of HTML5 to end up with the IETF, but > I still think it would make sense for feed readers to adhere to the > rules as well.) > > When sniffing was discussed a while ago I remember that > technorati.com and a feed library gsnedders was working on made > their code much stricter. They're not browsers. I can't find any reference to Technorati looking in the archives, but what SimplePie does (which is what I (sorta) work on) matches an old draft of what is in HTML 5 (the only differences don't effect our detection of text/html v. feed types from memory). What we did before that was do what Sam is describing, completely ignore the Content-Type header (that is actually untrue, as we used some regex like "; \s*charset=([^;]+)" to get a charset from it). Only a small number of feeds broke (all served as text/plain to my knowledge). -- Geoffrey Sneddon <http://gsnedders.com/> <http://simplepie.org/>
Received on Thursday, 28 May 2009 14:40:45 UTC