Re: Host headers for accelerator origins

Henrik Nordstrom wrote:
> On fre, 2008-07-18 at 12:21 +1200, Amos Jeffries wrote:
>> Henrik Nordstrom wrote:
>>> On tor, 2008-07-17 at 09:09 +0800, Adrian Chadd wrote:
>>>> Well, what we should really do is document exactly whats going on, so
>>>> we can then write tests to ensure the code matches the intent.
>>> The intent is that nothing goes on unless one explicitly asks for it.
>> As he did in this case with 'vhost' option.
>
> that tells Squid to use the Host header, which is the effect he saw.
>
> or did I misunderstand Mark completely?

I read his post as saying he configured with vhost then saw it behaving
like it was designed. But he misunderstood the intended option behavior
and wanted clarification.

>
>> My understanding was that the correct behavior is:
>>
>> - default or 'accel': no re-writing done. Assume the client did it
>> properly, too bad if not.
>
> accel alone is a no-go. Needs at least one of vhost or defaultsite
> telling squid how to pick up the requested hostname.
>
>> - never re-write to the cache_peer 'name' or label EVER.
>
> correct.
>
>> - with 'vhost' configure on the http_port should re-write URL domain
>> to the Host: FQDN.
>
> correct.
>
>> - with no Host: but defaultsite, should re-write as above to fixed
>> defaultsite value.
>
> correct.
>
>> - same for vport
>
> vport is a bit of an odd beast, but yes.
>
> Regards
> Henrik
>

Amos

-- 
Please use Squid 2.7.STABLE3 or 3.0.STABLE7