Re: draft-fielding-http-key-02 obvious shortcoming & failure

[Mark Nottingham <[e-mail]>]

Hi Henrik,

Yes, we're aware of that. 

We're trying to find the right level of complexity and capability for the draft, and biased towards "simple" very early. As we get some implementation experience and evolve the draft, we'll address this.

Cheers,


On Jul 25, 2013, at 8:39 PM, Henrik Nordström <[e-mail]> wrote:

> Learnt about the Key draft today in another discussion about how to
> cache Vary responses, and reading the document I see a noticeble
> shortcoming and failure of the proposed algorithm.
> 
> In "2.2.4. "p": Parameter Prefix Match Modifier" you have
> 
> Key: Accept;p="text/html"
> 
> And a seemingly nice looking list of things it matches and do not match.
> So far so good. But it will also match
> 
> Accept: text/plain;q=0
> 
> which is the opposite. Here the client says it do not accept text/plain.
> 
> same issue applies to any other header using quality attribute.
> 
> It also fails to represent quality selection among different variants in
> general. I.e. when there is both text/html and text/plain versions and
> one client prefers plain, the other html but both can process both.
> 
> Regards
> Henrik
> 

--
Mark Nottingham   http://www.mnot.net/