[prev in list] [next in list] [prev in thread] [next in thread] 

List:       squid-dev
Subject:    Re: Rationale behind storeKey?
From:       Mark Nottingham <mnot () yahoo-inc ! com>
Date:       2008-06-26 23:25:05
Message-ID: 5334223D-9465-4560-92F1-50ED70715E91 () yahoo-inc ! com
[Download RAW message or body]


On 27/06/2008, at 7:19 AM, Henrik Nordstrom wrote:

> On ons, 2008-06-25 at 10:32 +1000, Mark Nottingham wrote:
>> Question: why is the request method used as input to the storeKey?
>
>> What am I missing?
>
> The history of Squid, and in addition the negative cache of error
> responses.
>
> Caching of POST results (positive) should be done so that the the  
> result
> is cached as if the method was GET, when allowed.
>
> HEAD also should be special-cased more than it already is, but it's a
> bit dangerous to use it for updating cached headers as there is  
> numerous
> servers sending bad response headers in response to HEAD.. The main
> reason why HEAD is still a bit separate from GET is that we still can
> not cache partial responses.

Aha...


> A HEAD response is a extreme case of a
> partial response with 0 octets of the response body.. But the  
> intention
> is that there should only be a HEAD entry in the cache if there is no
> valid GET entry.


Makes sense. Thanks.


--
Mark Nottingham       mnot@yahoo-inc.com


[prev in list] [next in list] [prev in thread] [next in thread]