httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Graham Leggett <minf...@sharp.fm>
Subject Re: [PATCH]: mod_cache: don't store headers that will never be used
Date Sun, 29 Jul 2007 20:19:21 GMT
Joshua Slive wrote:

> What needs to be validated is if, in fact, the headers with NEVER be
> used. Is there any configuration directive or client request that
> could make mod_cache use headers from the cache when max-age=0? I
> don't think so.

The headers will be used by the next request - not just this one, and 
you have to be very careful if assuming anything about what the next 
request may or may not do.

What may make this workable is the combination of "The body is NOT 
stale" with "max-age=0". The danger of not writing the headers is that 
an entity, once stale, will not be freshened when the spec says it 
should, and will cause a thundering herd of conditional requests to a 
backend server. This issue has been badly understood by some in the 
past, who suggested that the ability to update cached entities be 
removed. We need to make very sure that by fixing one problem we don't 
introduce another.

That said there is a real issue of a frontend thundering herd problem 
caused by braindead clients, which inspired the patch in the first 
place. If this patch was made conditional with a directive, and not made 
default, then I think it could work.

I'm not sure - is there a directive that does "work around broken 
clients"? It may be worth adding one.

> Nah, I bet there are many times where the two sets of headers will
> differ (based on Date alone, for example, or Expires, which will get
> updated to match the max-age=0)

True.

Regards,
Graham
--

Mime
View raw message