Re: [RFC] remove negative_ttl directive ?

2013-01-11 Thread Henrik Nordström
fre 2013-01-11 klockan 18:53 +1300 skrev Amos Jeffries:

 Can anyone present any actually useful reason to keep it despite the 
 problems it presents?

referse proxies need to be able to cache error responses even if there
is no cache-control. But it does not need to be a specific directive.
Merging this functionality into refresh_pattern is fine.

Regards
Henrik




[RFC] remove negative_ttl directive ?

2013-01-10 Thread Amos Jeffries
The negative_ttl directive is continuously causing problems. What it 
does is DoS all clients of a proxy when one of them has a URL problem. 
In modern websites which can present per-client error responses targeted 
at an individual client this can be a major problem.


I propose dropping the directive entirely and following HTTP RFC 
guidelines about cacheability of 4xx-5xx responses.


The one case I can think of it actually being useful is to prevent DDoS 
against a reverse-proxy. However, due to DDoS usually varying the URL 
anyway this is an extremely weak protection.


Can anyone present any actually useful reason to keep it despite the 
problems it presents?



Amos



Re: [RFC] remove negative_ttl directive ?

2013-01-10 Thread Kinkie
Hi all,
  since we have it already what about taking a softer stance? For
instance, subject it to an hard-coded maximum limit of a few
seconds/minutes, or display a strong warning to cache.log when a value
above a certain time limit is found during parsing..

On Fri, Jan 11, 2013 at 6:53 AM, Amos Jeffries squ...@treenet.co.nz wrote:
 The negative_ttl directive is continuously causing problems. What it does is
 DoS all clients of a proxy when one of them has a URL problem. In modern
 websites which can present per-client error responses targeted at an
 individual client this can be a major problem.

 I propose dropping the directive entirely and following HTTP RFC guidelines
 about cacheability of 4xx-5xx responses.

 The one case I can think of it actually being useful is to prevent DDoS
 against a reverse-proxy. However, due to DDoS usually varying the URL anyway
 this is an extremely weak protection.

 Can anyone present any actually useful reason to keep it despite the
 problems it presents?


 Amos




-- 
/kinkie