Re: CRLF requirement

• From: Julian Reschke <julian.reschke@gmx.de>
• Date: Tue, 01 Jul 2014 17:54:11 +0200
• To: Anne van Kesteren <annevk@annevk.nl>, Carsten Bormann <cabo@tzi.org>
• CC: HTTP Working Group <ietf-http-wg@w3.org>
• Message-ID: <53B2D9A3.8020204@gmx.de>

On 2014-07-01 17:35, Anne van Kesteren wrote:
> On Tue, Jul 1, 2014 at 5:29 PM, Carsten Bormann <cabo@tzi.org> wrote:
>> On 01 Jul 2014, at 17:05, Anne van Kesteren <annevk@annevk.nl> wrote:
>>> That is not what happens in the most popular clients today.
>>
>> I’m still curious why you think that’s relevant.
>
> Network effects. Content will end up relying on quirks of dominant
> players, especially if those quirks are stable and interoperable among
> a class of clients such as browsers.

That's true when the "bug" is likely to happen, such as creating broken 
filename parameters in Content-Disposition by just using string 
concatenation (causing whitespace and non-ASCII to appear where they do 
not belong).

But here? Why would anybody do this, even by accident?

> ...

Best regards, Julian

Received on Tuesday, 1 July 2014 15:55:04 UTC