Re: cache statistics

[Mark Nottingham <mnot@akamai.com>]

I don't know that a 1.0 vs 1.1 breakdown will be useful; HTTP version
(as expressed in the request/status line) is hop-by-hop, while cookie
cacheability is implemented with a large degree of variation from the
specification, from what I've seen.

More to the point, a proxy has to guess how cacheable a cookie is; a
CDN can either dictate cookie cacheability, or let the content
provider describe it.



On Tue, Jan 09, 2001 at 01:47:40PM -0500, Fred Douglis wrote:
> >I took a look at an ISPs forward proxy cache stats from about 300GB of data an
> >d have found the following:
> >
> ...
> >A breakdown of the non cacheable stats shows:
> ...
> >Cookie in Response: 3.3%
> ...
> 
> 
> This is a hot button of mine, because I have seen this sort of 
> breakdown on numerous occasions.  Am I mistaken, or is a cookie
> in the *request* also *possibly* a reason why data might be 
> uncachable in HTTP/1.0, while for 1.1, explicit cache headers
> are to be used and the presence of a cookie is deemed irrelevant?
> 
> How did your data break down 1.0 v 1.1, and how many requests had
> cookies?  I say that the request is sufficient because a request that
> identifies a user can get a user-specific response even if the
> reply doesn't include a cookie.  However, I know there are many
> cases where caches choose to ignore that fact, and I recall hearing
> that some analysis showed that most content (at least images) will
> be the same regardless of the cookie.
> 
> Fred
> 

-- 
Mark Nottingham, Research Scientist
Akamai Technologies (San Mateo, CA)