W3C home > Mailing lists > Public > www-validator@w3.org > November 2007

Re: [Suggestion] Wrong environment variable specified when validating ($_SERVER['http-accept])

From: Sierk Bornemann <sierkb@gmx.de>
Date: Mon, 5 Nov 2007 19:33:45 +0100
Message-Id: <A8375ADB-8A89-4882-88B7-56C3DBBD134B@gmx.de>
Cc: David Coll <david.coll.78@gmail.com>
To: www-validator@w3.org


Am 05.11.2007 um 18:54 schrieb Andreas Prilop:

> On Mon, 5 Nov 2007, David Coll wrote:
>
>> And that the validator should specify it's accepted Mime-type(s)
>> to be more accurate,
>
> I don't see why. It should likewise not specify any Accept-Language  
> header.

Andreas, you know how controversely we have discussed that here on  
this mailing list and there on

http://www.w3.org/Bugs/Public/show_bug.cgi?id=785
and
http://www.w3.org/Bugs/Public/show_bug.cgi?id=848

> You want to send something to the validator and something different
> to the browser.

No, that is not what he wants, and that is not, what he tries to  
achieve with his solution.

> It is like sending a valid document to the validator but sending
> an invalid document to the browser - yet claiming "Valid XHTML".

No, it is like sending a valid document to the browser whith the most  
apropriate Mimetype that browser is able to understand.
If it understands "application/xhtml+xml", the browser gets  
"application/xhtml+xml", if the browser only understands "text/html",  
then the browser gets "text/html".
The validator is able to cope with "application/xhtml+xml".

The switches to distinguish and widely implemented in diverse  
programming and scripting languages as JSP and PHP, also with Apache  
configuration are widely tested and do what they are expected to.  
Since years. In real life. And not theoretically in a test bed. They  
all are based on the fact, that modern web browsers do send a talking  
accept header.
The Validator should behave as normal HTTP client, when requesting a  
web ressource from a web server. So, what on earth should prevent it  
to *not* send a reasonable accept header, like any other modern web  
browser out there, too?
If there is no reasonable argument to *not* send any accept header or  
to only send */* like IE does (which does mean "I accept really  
anything you feed me") and does renege, but there are *a few bunch*  
of arguments to provide a reasonable and talking accept header  
(wether providing it from validator's side or by tunneling it like  
Björn Höhrmann once proposed in http://www.w3.org/Bugs/Public/ 
show_bug.cgi?id=18#c1) in favor of a satisfying result of client- 
server dialogue.


Sierk Bornemann
-- 
Sierk Bornemann
email:            sierkb@gmx.de
WWW:              http://sierkbornemann.de/
Received on Monday, 5 November 2007 18:34:10 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 25 April 2012 12:14:26 GMT