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
