- From: Roy T. Fielding <fielding@gbiv.com>
- Date: Fri, 29 Feb 2008 12:52:09 -0800
- To: Ian Hickson <ian@hixie.ch>
- Cc: Julian Reschke <julian.reschke@gmx.de>, Geoffrey Sneddon <foolistbar@googlemail.com>, James Graham <jg307@cam.ac.uk>, Boris Zbarsky <bzbarsky@MIT.EDU>, Anne van Kesteren <annevk@opera.com>, "public-html@w3.org" <public-html@w3.org>
On Feb 29, 2008, at 12:17 PM, Ian Hickson wrote: > It's only necessary because the HTTP working group refuses to act in a > responsible manner and actually specify how to write an > interoperable user > agent that is both compatible with the Web and handles invalid > content. If > HTTP defined how to do this, we wouldn't be stuck with defining it > ourselves. Ian, you haven't the faintest idea of how to write a specification that will last beyond the current fad-of-the-day in implementations. If HTTP had been as foolish as HTML5 is being in regard to architectural decisions, the entire Web would currently be owned by one vendor (pick any one). All you are doing is ensuring that the future Web standard won't be called HTML because future platforms won't be able to implement this muck efficiently (unlike declarative HTML, which is inherently platform and implementation neutral). HTTP is an interface specification. Anything it might say about how to implement something that isn't defined by the interface would have to be applicable to all potential implementations (unlike your opinion that HTML5 is defined by what four browsers think they can implement tomorrow). In this case, it is far more interoperable for all implementations to explode upon seeing multiple CT headers, since that is a security issue even if the browsers react to it consistently. But I can't spec it that way because it isn't always practical to explode when the client is a small routine within a larger real-time system. All I can do is add a note to the security sections about handling errors, which is what the HTTPbis WG is going to do. Browsers are only one type of HTTP client and the only one that has persistent problems with implementation because some idiot decided that all browsers had to be bugwards compatible with XMosaic. The rest of the world actually expects software to report errors so that they can be fixed (and so that they don't become liable for misrepresenting lost data). Browsers don't do that because they are immature software systems, not because it is inherently better to hide errors. > You are right that we could have picked the FF/Safari behaviour > instead of > the IE/Opera behaviour, but you are wrong that it is ok for all the > browsers to do their own thing here. What we want is interoperability. That isn't interoperability. Interoperability is when the server and client agree on their interpretation of the message. What you are talking about is consistency in client implementation, whether or not it is interoperable with a given server. ....Roy
Received on Friday, 29 February 2008 20:52:28 UTC