Wikimedia Trying P2P Video Distribution
bigmammoth writes "One potential problem with campaigns and programs to increase video on Wikimedia sites is that video is many times more costly to distribute than text and images. The P2P-Next consortium has created an HTML5 streaming BitTorrent browser add-on to try and help experiment with ways to reduce the costs of video distribution. As described in a Wikimedia tech blog post, once the SwarmPlayer add-on is installed, and when using the multimedia beta, video on the site will be streamed via the hybrid HTTP / BitTorrent SwarmPlayer. For smooth playback the Swarmplayer downloads high priority pieces over HTTP while getting low priority bits from the BitTorrent swarm. The same technology is available for experimentation with any site via the standalone version of the Kaltura HTML5 Media library."
This is good news. It'll:
a) make it a lot easier to compete with the likes of youtube.
b) be very easy to take advantage of, once integrated into CMS's.
c) make it a lot harder to argue that P2P is only something that pirates use, rather than simply modern technology.
Why not just use Youtube to host the videos, after archiving them in a Wikimedia store?
Or is this more about control than openness? What value does hosting them at Wikimedia have over Youtube?
Youtube isn't that restrictive as long as you aren't infringing copyright..
The problem with slashdot is that most of its users were bullied and stuffed into lockers as kids!
It will make it much harder to use Wikipedia as a reference. You will want to look up something quickly and be presented with four our five possibly relevant 10-minute videocasts on the subject.
"It is unfortunately not available for your browser"
P2P is good for content which everybody wants right now but what about the situation where you have an encyclopedia full of videos and few of them are accessed by different browsers in any given day? Client side caches can't hold on to this stuff for ever. I wonder if there is any benefit from using P2P in this case.
http://michaelsmith.id.au
Isn't this a form of net neutrality, where we have high and low priority data ?
PLEASE SEED!!
This sounds nice. With the back up of HTTP server, this means the leechers will be irrelevant, when nobody else is seeding. Many people will just leech, watch and forget, without giving back seed. Because of the lack of seeds, the file hosting sites are so damn popular now. If everyone was seeding after download, then nobody would need file hosting services. Even YouTube, Vimeo and other video sites may go to this model then. (HTTP with backup of bit torrent and vice versa) cool cool cool. This gets even better for the people like me who have different Internet speed for country traffic and abroad traffic (I have 10 times faster traffic from local servers), so when torrenting from someone nearby, I will have faster speed, than downloading from the abroad server. Nice.
Are you thinking multicast? Because this is the real need here.
Can anyone fill us in on where is multicast on the internet right now? It seems pretty far away - even further away than IPv6.
Write boring code, not shiny code!
Lets see how the MAFIAA/Government/ISPs deal with P2P being used like this...
After years of not using a signature, I am going to make one to say the following: Fuck Beta
a obvious use for Bittorrent would be serve big backups of wikipedia as bittorrent.
but the backups of Wikipedia are not served that way, because make no sense, since at the speed that change, you will have people seeding a old version no one wants anymore.
bittorrent has not appeal for files that can change often.
just saying...
-Woof woof woof!
It's almost ironic, seeing as how Opera is the only mainstream* browser with a built-in Bittorrent client.
*For rather small values of mainstream
IPv6 includes multicast IIRC.
IPv4 multicast is basically broken by NAT, so is unlikely to ever get used on the internet itself.
The problem with multicast is that everyone is supposed to need the same packets at the same time, like in a real time TV feed. AFAIK it doesn't work so well when each person might want to start watching a video at any given time.
Dilbert RSS feed
The problem with multicast is that everyone is supposed to need the same packets at the same time, like in a real time TV feed. AFAIK it doesn't work so well when each person might want to start watching a video at any given time.
True, multicast isn't long tail friendly. But if a video gets linked from a Slashdot story, there will likely be enough viewers to justify starting a multicast swarm every so often. Send the first minute as unicast and the rest as a rolling multicast swarm. Then while a player is receiving the first minute over unicast, it can receive the next minute over multicast. Players on sufficiently fast connections might join multiple swarms at once to pull in the whole video more quickly.
How would it help if 90% of web users are behind NAT/proxy?
Who does have real IP on the desktop at all? Do I have to open inbound port directly to my browser?
I'd like to keep all P2P traffic on my router, so it doesn't get anywhere inside my LAN
But that requires an immense quantity of swarms (multiple per video), which means the routers will have to store huge in-memory databases to keep track of the downstream routers who need each packet stream, because unlike with unicast, the destination IPs are not contained in the packets themselves.
Dilbert RSS feed
Wikipedia - whose "fair use" justification is frequently "we couldn't find an image usable under the normal interpretations of fair use, so we used this one anyhow"
As far as I can tell, all non-free media on English Wikipedia, other than WMF logos, is supposed to be an excerpt (factors 3 and 4) used in context of commentary on the image's subject (factor 1), and the subject has to be of a nature that free images cannot be produced (factor 2). Most of these are of A. a notable non-free work of authorship or B. a notable person who is dead or extraordinarily reclusive. Can you cite specific abuses of fair use on Wikipedia so that I can file an IFD?
It's a wiki. Someone can upload a transcript. If you're lucky, it might even be in a timed text format such as LRC or SRT that can also serve as subtitles.
c) make it a lot harder to argue that P2P is only something that pirates use, rather than simply modern technology.
But... but... I thought that only evil pedo-terrorist pirates are using Torrent, to siphon the hard gained profits out of the pocket of the poor record- / motion picture- industry !~
More seriously : I'm actually surprised that it took so much time until someone decided to implemented it. Leveraging P2P to offload server load for user-made and -uploaded videos (just like it's already used to offload bandwidth requirement for distributed TV - like torrentocracy - and for upgrades - like in WoW. ) just make perfectly sense.
In fact, it would be good if some kind of standard was added, perhaps to HTML5.1 allowing other browser to participate in the P2P offloading in a standardized manner.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
it spawn a separate process that stays open and seeding as long as your computer is on and you don't close the application. It has a small indicator in the lower right of your browser and a system tray icon. If you want to 'turn it off' you can disable uploading which is recommend over complete removal since you can still help reduce server load by using the extension even if you upload nothing.
Google Gears also started as a plugin before being swallowed by HTML5.
Video where handled by plugins before being considered by the VIDEO tag of HTML5.
3D Web also started as a plugin before WebGL emerged as a standard.
see any tendency ?
Well, if things keep that way, bittorrent P2P server offloading could be integrated into HTML5.1 /. effect, if links on the main page where modified to leverage such swarming, in addition to direct HTTP access to the /.ed server.
could even help mitigating the
If enough proponents push for it (and it is really useful, so it's worth pushing for) we might start to see more widespread adotpion : Firefox, but also Chrome, Opera, Safari.
perhaps even one day IE. Eventually. In a distant future. Once they finally finish getting previous standards right....
The only shortcoming : hope that the SHA cryptographic hashing/digest algorithm used by bittorrent doesn't get cracked, or malicious users will start flooding swarms with bogus packet, trying to inject their own shock site- or vulnerability- laden data instead of the legitimate data the P2P network was supposed to offload (just like some are injecting bogus packets into eDonkey networks, as MD4 is not secure anymore).
The other short coming : having an easy access to configuration to enable/disable/throttle uploads, to avoid the outbound connections exceeding the limits of data plan. Better if it's automatic (like leveraging NetworkManager on Linux).
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
multicast is only really useful for live streams, not watch-on-demand videos.
$ unzip, strip, touch, finger, grep, mount, fsck, more, yes,fsck,fsck,fsck,umount, sleep
the routers will have to store huge in-memory databases to keep track of the downstream routers who need each packet stream
Congratulations. You've just explained why multicast was never deployed over the public IPv4 Internet even before NAT had become widespread.
You could also have worded it:
"Hey, instead of just sending you a packet addressed to every host, I'll send you a single meta-packet and you can figure out who needs it from there."
Multicast has advantages (less bandwidth usage), but at a cost of much beefier routers. IIRC, setting up routing of multicast traffic can be a pain.
SSC
I wrote my master thesis on a similar solution. I made a Python-based standalone implementation of the protocol and a simulator for it with everything happening on a single local machine. No nodes or no internet traffic. Kind of lame, but hey I got my degree :) Never bothered to do a proper implementation suitable for the real world use, as I got bored to tears by the time the thesis was done. Glad to see this something like this happening for real.
There goes another one of my ideas. Of course #bittorrent isn't logged. I should really start a darned blog :-(
it would be cool if they could use metalink, an internet standard for describing files offered in hybrid ftp/http/p2p content distribution systems, already used by a lot of open source projects.