Re: [squid-users] refresh patterns!

From: Adrian Chadd <adrian@dont-contact.us>
Date: Tue, 14 Aug 2007 14:22:21 +0800

On Tue, Aug 14, 2007, Manoj_Rajkarnikar wrote:

> >(Of course there's more to caching youtube - specifically, would need to
> >implement a patch to squid to create a URI from that youtube URL which
> >creates the same "host" part regardless of which bit of the CDN you fetch
> >it from - using that URL for the cache storage and lookup. That'd be a
> >pretty nifty start.)
>
> Ok so you mean its not that its not caching flashmedia but the url of the
> media file changes everytime you access it (request goes to different
> servers for same content)..

Well, it can change. Unfortunately the concept of a URI didn't "stick",
so there's multiple URLs which reference the same resource..

> 1187069957.907 803 202.51.76.26 TCP_MISS/303 276 GET
> http://youtube.com/get_video?video_id=69M_1ow_yEg&t=OEgsToPDskLjJ2R2yzfUrzuuPjSq4-2Z
> 1187069963.328 3857 202.51.76.26 TCP_MISS/302 181 GET
> http://cache.googlevideo.com/get_video?video_id=69M_1ow_yEg
> 1187069973.085 1574 202.51.76.26 TCP_MISS/200 431 GET
> http://video.google.com/s?ns=yt&sourceid=y&sdetail=p%3A%2F&vid=kPCRaxHXMKD2NSrRYUYeegC&docid=69M_1ow_yEg&el=detailpage&nbe=0&st=0.667&et=0.667&len=104&rt=14.7&fv=WIN%209%2C0%2C47%2C0

See, st and et and len and rt might refer to times, and therefore might be
some crazy way of handling range requests without doing range requests?

> Above is the request to same video twice. the part in the url "&t=.." in
> youtube.com/get_video?.. and "&st=.." and "&et=.." and "&rt=.." in
> video.google.com/s?.. keep changing with every request. but I do wonder
> why the second url cache.googlevideo.com/get_video?.. got the miss.

I haven't looked at google video at all. Youtube does seem nicer to try and
dream up rules to cache.

> anyways, we will not be caching any of these youtube
> and googlevideo urls for now as we donot have much big cache space and
> those flash videos would use up the cache space with minimal or none
> chance of getting a hit, IMO. Please correct me if i'm wrong in this
> assumption. But I'm very interested in caching these urls if someone could
> pull off a patch as adrian suggested (maybe rip off those above tags from
> the url while storing and looking up off the cache store). I'll be
> increasing the storage space in the near future so it would be great to
> see such a patch.. ;)

Yes, it'd be good if someone could come up with a patch to implement turning
a given URL into a "lookup URL" for cache storage and retrieval.
That'd at least fix the "youtube" style content caching. Google video probably
requires a little more thought, and maybe google just don't want
you caching their stuff.. (caching google maps stuff would be nice too..)

Adrian
Received on Tue Aug 14 2007 - 00:21:11 MDT

This archive was generated by hypermail pre-2.1.9 : Sat Sep 01 2007 - 12:00:03 MDT