[squid-users] time based range_offset_limit

[Amos Jeffries]

On 14/07/2016 12:43 a.m., Heiler Bemerguy wrote:
> 
> Em 12/07/2016 23:43, Alex Rousskov escreveu:
>>
>>> (without using Range: header).
>> That's your squid.conf customization, I presume.
>>
> 
> The squid won't send a Range: header to the server because the request
> is matching the range_offset_limit -1 ACL. I presume. So squid will try
> to fetch the file from the beginning, faking a full request, right?

No faking. Is making.

>>
>>> That's why I don't understand why it does not work on a REAL
>>> enviroment.
>> Many things can go wrong -- the real requests may require collapsed
>> forwarding that you do not test, the real requests may have no-cache,
>> the real response may not be cachable, or there is some Range handling
>> bug that your test scripts do not tickle (e.g., they request ranges that
>> are always close to each other and are always available at the same
>> time).
> Well, if I turn off collapsed_forwarding and try to GET the same file on
> the same server in a row (only changing the Range), it will create *two
> *connections to the server instead of only *one*.
> I use "override-expire ignore-private ignore-no-store ignore-reload
> ignore-must-revalidate store-stale" for this particular request, won't
> it override the no-cache or whatever?

No. Those refresh_pattern options are overriding the response
requirements mandated by the server.

The "no-cache" Alex speaks of is a client requirement that no cached
data be sent. Which also means that client request cannot be collapsed
with others, since collapsing is essentially just using 'cached' data
before it gets stored to the cache.

Amos