On 2 Sep 2010, at 18:20, Graham Leggett wrote:

> On 02 Sep 2010, at 7:01 PM, Nick Kew wrote:
> 
>> Indeed.  I guess my comment in STATUS was down to reviewing that
>> backport proposal (and checking the RFC) before I saw the other one.
>> 
>> I guess the real question is: why enable it in the abstract, in the
>> absence of a backend implementation?  Surely there's no backend
>> that's going to be simpler to support ranges in than plain ol' disk?
> 
> mod_cache.h exports a public provider interface, allowing external parties to 
> write implementations of their own that are distributed separately from the 
> httpd server.
> 
> It is broken for mod_cache to dictate to an external implementation that it 
> is forbidden from caching partial responses, when the caching of partial 
> responses is allowed by RFC2616.

I disagree about 'broken': a cache isn't *required* to cache ranges.

As for third-party backends, this change is at serious risk of breaking
anything that has (perfectly reasonably) ignored range responses.
Fine for trunk, but not for a stable branch.

-- 
Nick Kew

Reply via email to