Bittorrent implementation

Message from John Cave

I would like to start a bit of a brain storm into how we implement bittorrent, because it is something we have not discussed during any of our meetings.

Most distros offer an official torrent. It is a great way to deliver the distro without using up too much of our servers' resources. We can say "we would prefer you to download the torrent if at all possible" and many people will use it. Furthermore, it is an easy way for people to give something back to the distro, by allowing them to upload the file and help to spread it.

So, the questions are this:

1. Will we set up a separate CMS for the .torrent files or simply provide a page with the links to them? An example of the CMS route is Sabayon http://tracker.sabayon.org/ The CMS would not require any login (whew).

2. What will we use as the central tracker? Will we set up our own or use something like publicbt? I know setting up our own sounds very silly, but using our own does cut out any association to the darker side of bittorrent.

3. Will we set up a small virtual machine (for example with only 256 mb of RAM) to seed the torrent very slowly? Only one hundred kilobytes or even fifty would suffice - just to make sure the torrents never stall.
   My own server would be seeding pretty much permanently - but I couldn't guarantee uptime and it would be very very slow (~6 kb/s). I know how annoying it is to start a download, get to 89.2% and have it stop. I am doing that right now, actually. Is there some way we could automate the process of making the .torrent or would we have to make one ourselves every year?

My suggestion is to have a CMS, use a public tracker, have a server to seed it slowly and automate the process if at all possible.