Re: [squid-users] passthrough download manager's multipart downloads / range requests?

Hi Amos,

thank you very much for taking the time to help me out!

I had no idea that this was supposed to work by default.

This explains why I wasn't able to find anything :D

I used the default "privacy" config (explicitly allowing some
request_headers and deny the rest) from the wiki.
The problem was caused by not allowing the "Range" request header
("Range-Request" was allowed already).

Now it works :)

> NOTE: download managers which open parallel connections are *degrading*
> the TCP congestion controls and reducing available network resources
> across the Internet. Reducing their parallel requests to a single fetch
> is actually a good thing.
>

Understood.
Using this squid proxy for my tiny home network it's just that it can
get really annoying when I have to download a single file only and the
maximum I get with 1 connection is a few kb/s. So I'm unable to resist
the temptation ;)

Thanks again for all the info. I was on the wrong trail all the time.

Kind regards,

Max

On 01/01/14 06:27, Amos Jeffries wrote:> On 1/01/2014 12:15 a.m.,
mxx_at_mailspot.at wrote:
>> Hi,
>>
>> Maybe because most of the time squid is used differently I'm having
>> troubles finding an answer to this question.
>> It would be very nice if someone could help me out with this :)
>>
>> I only use it to filter ads and to redirect traffic to some domains
>> through different uplinks. I don't really need the caching.
>>
>> Squid 3.4 does all of that perfectly (Linux 3.12) in intercept mode.
>> But download managers using multiple connections concurrently to
>> download 1 file are only able to use 1 connection/destination anymore.
>
> Squid does not impose any such limitation. Unless you have explicitly
> configured the maxcon ACL to prohibit >1 connections per client IP.
>
> The default behaviour of a non-caching Squid should be exactly what you
> are requesting to happen.
>

>>
>> What I've found so far are only options like range_offset_limit in
>> regards to cache management.
>
> If you have configured that range limit or the related *_abort settings
> then they may cause a behaviour similar to what you describe. Not
> exactly a prohibition, but Squid downloading the entire object from
> start until the requested range is reached. Doing that N times in
> parallel can slow down the 2..N+1 transactions until they appear to be
> one-at-a-time occurances.
>
>>
>> Is it possible in any way to let squid pass through and simply ignore
>> all connection requests to destinations with certain Content-Types so a
>> client could connect multiple times to the destination concurrently?
>
> Content-Type is not known until after the request has been made and
> reply received back. What you ask is like deciding whether to make an
> investment now based on next years stock exchange prices (the URL can
> give hints of likelihood, but is not very reliable).
>
> Amos
>
Received on Thu Jan 02 2014 - 21:38:59 MST