Re: Content Sniffing impact on HTTPbis - #155

Ian Hickson wrote:
> On Sun, 14 Jun 2009, Adrien de Croy wrote:
>> If there were no distinction between "resource" and the bag of bits 
>> returned, how would one go about multi language content, where the URIs 
>> are the same, but different bags of bits are returned for the same URI 
>> requested based on Accept-Language?
> 
> Different resources are returned every day for the same URI. For example, 
> news.google.com differs from minute to minute. Two different people 
> looking at the same Google search results page get different results 
> depending on their search history. Why is that a problem? When you access 
> a server and give it a particular URI, you get a resource back. No need to 
> have a nebulous theoretical concept of "resource" that doesn't actually 
> exist; it isn't needed to explain what's going on, and most people -- even 
> most people who work with Web technologies daily -- don't think that way.

*You* may not need that distinction. However, when writing a spec, you 
aren't doing it for yourself, but for an audience of readers.

Other spec writers and standards bodies *clearly* see this differently. 
Changing terminology at this point is only going to cause confusion.

So, as Mark said, you can ignore the advice, and then will have to 
change things anyway once you attempt to publish as RFC.

> (BTW, can you point me to a site that actually does that with Accept- 
> Languge? Or any HTTP con neg for that matter. I've never been able to find 
> any site that does con neg other than demos, experiments, and user agent 
> tests. I'd love to actually test how browsers do this in real-world 
> situations rather than in controlled experiments.)

I have written servers that use Accept-Language when the authenticated 
user's locale was unknown. But I currently can't point to any public 
instance you might be able to test with.

Other than Accept-Language, there is an incredible amount of conneg 
going on; consider "Accept-Encoding:", "Cookie:", and "User-Agent:" -- 
yes, all of these cause content negotiation once the server decides to 
vary the representations it sends.

BR, Julian

Received on Sunday, 14 June 2009 06:55:43 UTC