Slashdot Mirror
Bittorrent To Replace Standard Downloads?
Max Sayre writes "Have you ever tried to download an operating system update only to have it fail and have to start all over? What about patches for your favorite games? World of Warcraft already uses Bittorrent technology as a way to distribute large amounts of content at a lower cost to the company and faster speeds to all of their clients. So why haven't they replaced the standard downloading options built into any major OS? Companies like Opera are including the downloading of torrents in their products already and extensions have been written for Firefox to download torrents in-browser. Every day Bittorrent traffic is growing. Sites like OpenBittorrent already exist and DHT doesn't even require a tracker. So why isn't everyone doing it? Is it finally time to see all downloads replaced with Bittorrent?"
When torrent support comes equipped on all the major browsers, it can take off.
Until then it's a tool for nerds to get their porn faster.
Mod me down, my New Earth Global Warmingist friends!
Combine this with social networking to allow/deny access to your files and I think you've got a game changer. Files which require no server, and which are unknown/unavailable to anyone who doesn't need to know about them. I could share my mp3 collection or movie collection with only my friends list, which would be much more along the lines of fair use (like tape trading).
Why? because for small files (as I expect most software updates would be), downloading directly is quicker and safer.
Ask me about repetitive DNA
Those of us stuck in New Zealand or Australia still have data caps to think about. If every download was a torrent there would be a lot more overhead eating into our precious data caps!
Please, think of the Kiwis.
Because Bittorrent has a reputation issue, for one. The MPAA and RIAA attack it and call it the reason they are losing money (instead of their failing business model).
Large companies don't want to have to deal with the previous hassle, and even though the load might not be much for individual computers, if everyone on a company network was bittorrenting, other traffic would be interrupted (even on 2MB DSL, bittorrent interferes with my connections to many popular IM services and I don't even run it full throttle during the day).
"There is a way that seems right to a man, but its end is the way of death." Proverbs 16:25 (NKJV)
1) because I'm a leech.
2) because I don't want legal liability FOR DISTRIBUTING if I download a file that unknown to me is illegal, e.g. a software package from overseas that someone inserted illegal-in-my-country pornography into the binary. Yeah, I'll take the risk for possession but not for distribution.
3) because my employer's lawyer made me say #2 when it comes to company machines.
4) because I prefer to get my bits from the official location. Yea, I know a checksum should be good enough but I'm old school here.
Seriously though, I can see torrents overtaking web- and ftp- downloads as the primary method for distributing large, popular files. However, there will always be customers who refuse to share and who refuse to get data from any source that doesn't have a reputation for quality and isn't blessed by the original publisher.
Oh, and seriously, I'll be fine using torrents to download things like well-known linux distros. I trust modern checksums. I probably won't use them for low-demand files or smaller files though.
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
There was apt-torrent, but that project appears to be abandoned.
The thing is probably that there is no pressing need. There are many educational facilities that are are willing to provide mirrors for such things, so there's no real reason to implement a system to borrow user's upstream bandwidth.
upon the advice of my lawyer, i have no sig at this time
Because for security updates, this allows users to find others who don't have the latest patches yet. Just imagine the people watching leecher IPs every time a new remote exploit is patched...
In WoW I have to disable bittorrent if I actually want to download a patch. Otherwise it saturates my connection with upload data whilst only downloading at 1% of my max speed.
Blizzard use bittorrent simply because they're cheap. Instead of using their millions in profits to provide bandwidth, they make the players smash their quotas sending data to each other. I had to install a bandwidth limiter to get Wrath of the Lich King to install because otherwise the outrageous upload speeds stopped me actually downloading anything. You'd think $15 a month would be enough to pay for enough bandwidth to allow me to download the game I've just paid for, but no they have to chase every penny...
I would like a solution that combines metalinks (one file that contains multiple urls for a download plus the checksum) with Bittorrent.
A client could start a http download from one server, and a bittorrent that requests pieces for the latter chunks. You can also make multiple http request with a offset these days, on another http server or the same one.
This could even be built in magically into http browsers: if the file size is > 50MB, ask the cloud if there are nodes for the given url. That is provided you have a checksum like with metalinks. Appearantly metalink already features this possibility: http://www.metalinker.org/
NB: The message above might reflect my opinion right now, but not necessarily tomorrow or next year.
Most modern bittorrent client support web seeds, that is using an http-hosted file as a seed for the torrent. Ad the speed from that server to the other people who are downloading and you have much better speeds than if you were to simply download straight from the server. Add to this all the other bittorrent features, like resuming a broken download, and improved error checking and you have a very powerful downloading strategy. Just take a look at burnbit: http://burnbit.com/ which takes a normal hosted file on the internet and turns it into a torrent. Everyone wins!
The file sizes of most Linux packages are simply not big enough to warrant the use of bittorrent. The 32 bit x86 kernel(usually one of the biggest packages in a distro) is only about 32 megs or so. By the time you downloaded the tracker, found your peers and actually started downloading something you could have had the whole package d/led already. Most big universities and research institutions have to host the files anyway(for internal updates), its not all that difficult to extend the download service to the general public. Not to mention the fact that in order for the torrent to be effective you would actually have to retain the packages after installation which can quickly become a huge pain in the ass.....
Monstar L
Bittorrent is great for very large files, and popular files.
But for small files it's really, really bad. Many linux patches involve downloading hundreds of small files, not one big one. Most applications are so small that the setup and teardown time for bittorrent would dwarf the download time. Any download that takes less than 5 will likely have a smoother user experience if it is not done using bittorrent.
Even ignoring tiny files, there is the issue of bandwidth limited users, the significantly higher routing requirements of bittorrent (many home routers flake out when you get 50+ TCP connections going through them), users with heavily asymmetrical connections (5Mbit down/256kbit up), and the more complicated configuration required to get a good bittorrent connection.
In short, bittorrent is nice for its niche (large, popular files), but outside that niche it is often not the best solution. Wider deployment of bittorrent technology would probably help some places, but it's not a silver bullet for all Internet downloads.
"I will trust Google to 'do no evil' until the founders no longer run it." Hello Alphabet.
Contacting the tracker and getting an initial peer list, in a proper system, takes a fraction of a second. It's iteratively contacting peers, obtaining their piece bitmap, negotiating with them for piece exchange, and finding peers that actually have high bandwidth that makes the startup time of BitTorrent so high.
Say you've got a CD or DVD that's scratched, or an .iso you spent forever downloading via ftp and discovered to your dismay was corrupted. Assuming a bit-identical image is available online via .torrent, you can 'repair' your data without having to download the whole thing all over again:
Start your bittorrent app and begin downloading a new copy of the image you need. Immediately stop the download and exit your bittorrent app. An .iso file (incomplete, of course) will have been created in the destination folder.
Now rip your [damaged] disc to hard drive, creating an [obviously corrupted] .iso. Copy/paste that .iso into bittorrent's download folder, overwriting the existing .iso.
Fire-up bittorrent and begin your download once again. Bittorrent will analyze the corrupted .iso and immediately download the bits needed to repair (i.e. complete) it. In most cases this will only take a few seconds, even over dial-up, due to the insignificant amount of data usually needed (except, of course, in the event of a heavily scratched disc, which can also take a long time to rip in the first place; having a high-quality optical drive with good firmware and good optics certainly couldn't hurt).
This is spot on.
WoW (and other MMOs) needs torrents, because they have an very high to extremely high burst ratio. When a new patch is deployed for Linux it needs to propagate out to various distributions, people need to start packing it and then end users auto update will pick it up eventually. This means it's often distributed over time. When WoW deploys a new patch, they have 10 million people trying to get it at once in order to be first in new instances. Even the big distributors have trouble coping with this - and over a short period of time the need for bandwith will drop to very low levels as people are getting up-to-date, so there is no financial incentive for Blizzard to invest in the hardware to cope with deployment.
But if the repositories were themselves seeding, then it'd work just fine: Worst case is that it's still at least as fast as HTTP or FTP from the same repository (plus or minus some BT overhead), all else being the same.
Best case is that there's several repositories all seeding the same basic set of random apps, plus a bunch of users who have already downloaded the random app, and things turn both faster and cheaper than they otherwise would have been.
The hash checks performed by BT will do well to prevent errors and/or poisoned apps, as well.
Sounds like a win to me.
Kid-proof tablet..
Why has no one mentioned this?
Because multi-cast doesn't work in practice.
Because almost every gateway router drops multi-cast packets.
Because multi-cast is only efficient if there is more than one recipient on the same subnet downloading the file at the same time.
Because synchronizing the assembly of milti-cast downloads that were initiated at different times (as in, 1 second apart) would require as much work as implementing the bittorent protocol.
Honestly, multi-cast was only really thought out for machines sharing a private network. It wasn't intended for internet-style applications.
Yeah, but the built in download manager for Firefox sucks. Yes, it *can* resume... if you paused the download first. Even when it's told the file size, if your connection dies in the middle (which my connection too often does), it thinks it's "done" (which is absurd, because wget reports an error code) and you can't resume from within Firefox any more. Oh! I got an RST packet, that must mean that the last 300 MB transferred instantly!
I gave up on that piece of crap and used an add-on to make Firefox use wget, which at least has the decency to know when files have *actually* been fully downloaded, rather than giving up and deciding it's good enough to hand me a useless half file and no error messages at all. Seriously, have the devs never tested large downloads on a link that dies? It's not even hard to test: you can emulate the connection dying by pulling the damn ethernet cable. It just ignores the error and continues blindly. Does it really think that the other server going silent is an indicator that the file has been fully downloaded, or what?