Re: Re-registration of text/html

On 11.03.2010 11:01, Henri Sivonen wrote:
> "Julian Reschke"<julian.reschke@gmx.de>  wrote:
>
>> On 10.03.2010 16:33, Anne van Kesteren wrote:
>>> On Wed, 10 Mar 2010 16:24:59 +0100, Julian Reschke
>>> <julian.reschke@gmx.de>  wrote:
>>>> On 10.03.2010 16:12, Henri Sivonen wrote:
>>>>> This philosophical question could be avoided by stating that
>>>>> documents labeled "text/html" must be processed according to
>> HTML5.
>>>>> ...
>>>>
>>>> If we did that, we'd have to make sure that this doesn't break
>>>> existing uses. So statements like
>>>>
>>>> "User agents should ignore the profile content attribute on head
>>>> elements."
>>>>
>>>> would need to be fixed.
>>>
>>> Isn't that statement true for the majority of existing usage of the
>>> profile attribute? And therefore a SHOULD requirement is adequate?
>>
>> Why would a SHOULD requirement be adequate in that case? Me confused.
>
> You are allowed to violate a SHOULD if you have a good reason to deviate from the default requirements. If you have non-browser legacy software that doesn't ignore @profile on content crafted specifically for that kind of minority legacy software, presumably keeping it functioning as before is a good reason.
>
> In the general case, it's still good advice to say that UAs SHOULD ignore @profile, since Existing Content using @profile has been created in a context where the virtually all text/html consumers ignore it, so continuing to ignore it maintains legacy-compatible behavior.

That doesn't make sense to me.

Profile always was an opt-in switch. Either you recognize a profile URI 
and act on it, or you don't. If you don't know about any specific one, 
ignore them all. That has always been the case, and continues to be the 
case.

That's totally different from "SHOULD ignore".

Best regards, Julian

Received on Thursday, 11 March 2010 10:41:07 UTC