[squid-users] http://bugs.squid-cache.org/show_bug.cgi?id=4223

[Amos Jeffries]

On 27/01/2016 1:03 p.m., Tory M Blue wrote:
> Can we get an update on the bug mentioned here "
> http://bugs.squid-cache.org/show_bug.cgi?id=4223"
> 
> With this unfixed one can't use siblings with HTCP or  anything actually. I
> should be able to have my origin and  a sibling, I should be able to make a
> request to my sibling for a document and if that fails the request goes to
> the origin, and not pass back the failure from the sibling.
> 
> Just wondered why this bug is allowed to persist?

Good question. Nobody with money has yet been bothered enough to pay for
it to be fixed. Those of us doing Squid code as a hobby in our spare
time are currently focussed on either other bugs and other components
totally unrelated to it.


FYI: the quick workaround would be a config option added to cache_peer
to disable sending the "Cache-Control:only-if-cached" header on requests
to that peer.


The full long-term fix would be to also do an audit of the code handling
the sibling response to see why 500 is not doing exactly as you
described above.

Amos