On 18/11/2011 1:26 a.m., Ghassan Gharabli wrote:
I dont think that Squid2.7 could cache thing content unless it is
fully Http/1.1 supported but is there any work around to force caching
such a content because this URL is causing a big delay when the
startup of Youtube Video happens as Youtube Videos are being cached
then you wait about 7seconds to 9 seconds waiting till Youtube Video
starts to load.
YouTube is always changing its content& the way they allow their
partner video provider such a company to show a "LOGO" of their own.
When you log its content then you notice few output video url like
(*.youtube.com/generate_204) , (*.youtube.com/get_video) or
(*.youtube.com/videoplayback) and sometimes the ID can be seen at the
end of URL or sometimes can be seen in the middle of the URL and some
of the remaining few videos you can see at the start of URL after
"videoplayback" .. wondering why I cant see such information on
http://wiki.squid-cache.org/ConfigExamples/DynamicContent/YouTube ..
maybe it is not updated yet ?.
Maybe. We depend on people wih interest such as yourself to notice and
update the wiki.
I think the logo stuff is not mentioned because its relatively new, and
not actually relevant to the storage of the large video objects. The way
YT do it is actually cache friendly in that it allows the video to be
cached separate and the adverts left dynamically assembled at view time.
As for the ID; the patterns in Chudy's script are designed with ".*"
pattern pieces to pull the ID out of any position in the video URL path.
Dropping any surrounding fluff in the path parameters. Depending only on
the "itag=" being before the "id=" tag.
Caching this could present clients with a lie,
indicating that some server state has been changed when the server has not
even been contacted.
I thought of caching this content
http://www.youtube-nocookie.com/gen_204?attributionpartner=vevo just
to reduce the big latency even if the video was cached. I found it
interesting to cache the whole videos at Youtube thats why I preffer
presenting clients with a lie avoiding the big latency.
The key thing being that there is _no_ content to cache.
"Content-Length: 0" the other URLs that seem to come back from that
domain have per-user account details fairly often. It seems to me not
safe to cache it. Although safety has not been a strong barrier to
everybody, and they do not quite go so far as to mark it fully private.
You might want to experiment with a full block of the requests. That is
always safe to do but may prevent viewing.
Amos
