W3C home > Mailing lists > Public > public-webapi@w3.org > May 2008

Re: XHR LC comments

From: Maciej Stachowiak <mjs@apple.com>
Date: Fri, 16 May 2008 01:53:15 -0700
Message-Id: <C9F02C75-13F0-4984-BB75-09D89A9BF53A@apple.com>
To: Julian Reschke <julian.reschke@gmx.de>
Cc: "Web API WG (public)" <public-webapi@w3.org>


On May 16, 2008, at 12:04 AM, Julian Reschke wrote:

> Maciej Stachowiak wrote:
>> In practice it is much more important for same-origin to be  
>> implemented
>> consistently between XHR and HTML5 (and other Web standards) than  
>> for it
>> to be precisely consistent cross-browser, as inconsistencies in the
>> same-origin policy could lead to security holes. Thus, taking a  
>> snapshot
>> of what HTML5 says and putting it in XHR1 would be a dead letter,
>> because if HTML5 changes and browsers change to match it, they will  
>> not
>> leave their XHR implementation using an older version of the security
>> policy.
>
> Interesting enough, this seems to be exactly the opposite of what Ian
> just said :-):

HTML5 and browsers all differ slightly from each other on these  
issues. Though HTML5 does not aim to invent anything in the area of  
cross-domain security, I think there will be iterative convergence  
among the implementations and the spec.

The point is, if XHR1 ends up requiring something different than HTML5  
does, at least one of those will be ignored by implementors. Or to  
look at it another way, either HTML5 will not change on anything it  
requires on this, in which case citing its definitions won't actually  
change the meaning of XHR1 in the future; or it will change, in which  
case having an obsolete copy of the definitions in XHR1 will be  
actively harmful.

So we should either cite by reference or be prepared to promptly issue  
errata in the future.

Regards,
Maciej

>
>
> Ian> The point is that Apple and Microsoft are both going to  
> implement the
> Ian> thing as required by the Web in 2000, not as defined in HTML5.
> HTML5 is
> Ian> describing existing practice on these matters, not defining new
> material.
>
> BR, Julian
>
>
>
>
Received on Friday, 16 May 2008 08:54:02 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 16 May 2008 08:54:02 GMT