Re: Proposal: Vary on Cookies

On Thu, 12 Jun 1997, Richard Vermillion wrote:

> Dave,
> 
> Thanks for your response.  I did take a look at the Transparent
> Content Negotiation draft, but it didn't seem to address my problem
> directly.  Correct me if I'm wrong, but draft-ietf-http-negotiation-02.txt
> seems to discuss only the normal usage of Vary, i.e. using it with actual
> complete headers.  In Section 10.6.1, it discusses contructing "elaborate
> Vary headers", but these are constructed using only the Negotiate, Accept,
> Accept-Language, and Accept-Feature headers, e.g.:
> 
>     Vary: negotiate, accept, accept-language
> 
> I see your point in that I am basically talking about doing content
> negotiation with a remote variant selection algorithm which happens to
> take into account cookie values.  The unfortunate thing is that all of
> the cookie values are lumped into one header, "Cookie".  So, I'd like
> to be able to tell compliant proxy-caches that I only "selected" on
> a part of that cookie header.

Using cookie values would be my last choice. Getting agreement on an
approach to have proxies unravel cookies to examine individual values
is full of problems considering the difficulties we are having getting
agreement on a workable cookie standard.

But a number of values
you represented in cookies might also have been represented discretely as
other user attributes ... if the UA cared.

Also, the current microsoft 'position' would seem to be that a bunch
of your cookie based values would actually be negotiated by JScript 
within the browser ... as I understand the approach, an improved browser
object model plus existing JScript cookie access would cover issues by
letting the JScript decide which content to obtain, etc.

I think a combination of use of UA attributes for TCN and JScript for
client side dynamic content will come awfully close to what you want to
achieve.

Dave Morris

Received on Friday, 13 June 1997 01:26:52 UTC