Re: [squid-users] Youtube Squid

From: Amos Jeffries <squid3_at_treenet.co.nz>
Date: Thu, 19 Aug 2010 05:40:36 +0000

On Thu, 19 Aug 2010 10:47:30 +0600, Nyamul Hassan <mnhassan_at_usa.net>
wrote:
> Please read the following document:
>
> http://wiki.squid-cache.org/ConfigExamples/DynamicContent/YouTube
>
> That has some information on why caching YouTube is so complex.
>
> Somebody else will have to reply on the bounty.
>
> Regards
> HASSAN
>
>
>
> On Thu, Aug 19, 2010 at 10:28, Depo Catcher <depocatcher_at_gmail.com>
wrote:
>>
>>
>> On 8/10/2010 12:15 AM, Amos Jeffries wrote:
>>>
>>> Rafael Gomes wrote:
>>>>
>>>> Is there any other option to make it in Squid3 ?
>>>>
>>>
>>> No sorry. That feature is still on the list of things waiting for
>>> somebody to have the time and/or sponsorship to do.
>>>
>>> Amos
>>
>> Newbie question here, why is youtube different than other web content?
>> Because it's streamed?
>> Does this happen with all streaming data or just specific to youtube?
>> Is there not a standard way to stream data that is proxy friendly?

No, the problem as that page HASSAN referenced explains is their URL
design. It's intentionally adding obfuscation values with apparently no
other purpose than to break caching. Attempts have been made at least twice
to get them to be friendly and have fallen through.

Other content delivery sites have similar design flaws with the same
effect. YouTubes popularity expands the relatively small problems into a
major ones.

>>
>> What would sponsorship cost? Is there a chip in page or something that
>> we could donate to?

Cost is the work time for some programmer to port the Squid-2.7 storeurl_*
feature patches we have into Squid-3.HEAD code and test that they work.
Preferably also to work on some of the bugs found in the 2.x
implementation. I'm happy to provide a list of the 2.x patches out for
anyone assisting with this porting.

Amos
Received on Thu Aug 19 2010 - 05:40:49 MDT

This archive was generated by hypermail 2.2.0 : Thu Aug 19 2010 - 12:00:02 MDT