Re: Approved TAG finding: Authoritative Metadata

On Aug 8, 2006, at 8:16 PM, Ian Hickson wrote:
> On Tue, 8 Aug 2006, Roy T. Fielding wrote:
>>
>> On Aug 8, 2006, at 6:14 PM, Ian Hickson wrote:
>>> This isn't because of lazyness. This is because ANY BROWSER THAT  
>>> ACTUALLY
>>> TRIES TO IMPLEMENT THESE THINGS WOULD LOSE ALL MARKET SHARE.
>>
>> No, that's total speculation.  None of them have even tried to  
>> implement
>> a configuration option for identifying incorrect content types as an
>> error, let alone deployed it in the market.
>
> Actually, they have (and I've been directly involved with their
> development, usually in a QA capacity). They never make it past the  
> beta
> release stage because the vendors get so many bug reports and  
> complaints
> about sites breaking -- critically important sites like cnn.com,
> slashdot.org, hotmail.com, myspace.com -- that they have to revert  
> to the
> sniffing behaviour.

Ian, the finding calls for the feature to be configurable by the user
and does not require a specific default.  The feature could be as simple
as placing a little smiley/frowny icon in the corner of a window that
indicates whether or not the content is in error.  If a site "breaks"
because of such behavior, then the browser implemented it wrong.

> This isn't speculation at all. I'm speaking from direct personal
> experience working for and with browser vendors.

And I've been testing their products since 1993.  This has nothing to
do with the configurability of Web servers -- there are at least six
different ways to configure custom types on 97% of the deployed server
platforms.  The "wrong type" problem wasn't even an issue until MSIE
was deployed and sniffed data by default, after which authors would
neglect to set the media type because they simply never see the problem
themselves.  Sniffing data makes it far more difficult to deploy new
types because authors cannot change browser behavior.

>> It exists so that the same data format can be interpreted in  
>> different
>> ways, depending on the nature of the resource.  There is no way  
>> that you
>> can accomplish that with sniffing, just as there is no way that  
>> HTML5 is
>> going to change the normative interpretation of an IETF standard.
>
> With all due respect, standards are of no use when they aren't  
> followed.

Standards tell people the right way to do something for everyone
concerned. If people choose to do things the wrong way for their
own selfish reasons, then they are liable for the result.  Standards
cannot define irresponsible behavior as the norm and expect the rest
of the system to work.  Eventually, one of the browser vendors will
stop copying every mistake made by the others and let the market
decide for itself.

....Roy

Received on Wednesday, 9 August 2006 04:19:44 UTC