User:LFaraone/Ideas/BitTorrent
< User:LFaraone | Ideas
Revision as of 11:25, 19 December 2009 by Vd9 (talk | contribs) (User:LFaraone/Ideas/BitTorrent moved to Walter's rectum 25/Ideas/BitTorrent: Walter's rectum)
What:
- BitTorrent in Browse and on the wiki
Why:
- OLPC's servers will be put under considerable stress due to large amounts of users visiting the wiki from the third world. As the sole provider of activity downloads, OLPC will be serving files that can be very large; some of the Wikislices are larger than 100MB, not to mention the 300+MB OS images.
- BitTorrent is an open protocol, and there are many open source implementations, many of which are written in Python or have Python bindings. Integration of existing libraries allows for lower maintenance costs and less up-front-effort.
- BitTorrent is efficient; it requires only a server to "track" a torrent (and some torrents are trackerless). The server has minimal work, other than the initial "seeding" of a file to the public. As torrents are more popular, the speed often increases, as each new downloader is also increasing the "health" of the torrent via uploading.
How:
- Browse will handle a bittorrent just like a nomal file, except it will download the torrent file and pass it to the journal, which will execute the real download.
It is undecided how long it should continue to seed after a download has completed.
