Comments on: Public goods from public agencies

By: Stephen Bounds

Stephen Bounds — Tue, 30 Nov 1999 00:00:00 +0000

Nicholas,

The actual hard disk space required to store a substantial chunk of the ABC’s archives wouldn’t be super-expensive.

However, that’s predominantly a fixed cost. The killer is the *bandwidth* required to make the downloads available to all and sundry.

Let’s say 50,000 users access the service and each download two shows a week, on average, of 10MB each (a fairly conservative estimate).

That’s around 4,000 GB a month of downloads. Assuming bandwidth costs of $5/GB and excluding other maintenance costs including hiring systems administrators and so on, this modest system would cost around $240,000 — hardly chump change for the ABC.

By: Nicholas Gruen

Nicholas Gruen — Tue, 30 Nov 1999 00:00:00 +0000

Yes, but a massive amount of the bandwidth relates to recent programs. I’d be surprised if having the rest available as an archive adds much to the monthly bandwidth. Also, if one distributed all files as MP3 files, then they could be secondarily distributed through P2P systems. One could then choke the bandwidth of all files older than - say 4 weeks - but leave ID codes for each program on the website. Then you could search the file on a P2P file sharing community.

Then again, given my ignorance, there could easily be something wrong with this logic. Any suggestions?

By: Stephen Bounds

Stephen Bounds — Tue, 30 Nov 1999 00:00:00 +0000

P2P is an interesting idea. However, I think if the ABC were to choke off provision of the files after a certain time period accessibility to files would be greatly compromised. Studies have repeatedly shown that a majority of P2P users are consumers, not producers; most people do not make their downloaded files available to others for upload, meaning that after 4 weeks there may not be many copies in the community to find.

Additionally most companies, public or not, tend to dislike losing control when distributing their works and the ABC may still be suspicious of such a decentralized concept.

This makes the situation a perfect scenario for using a P2P program like BitTorrent (www.bittorrent.com).

BitTorrent differs from other P2P programs because it uses a central ‘tracker’ URL serving .torrent files to reliably identify the file to download. This tracker information is then used to identify other users currently downloading this file and efficiently obtains a copy by sharing and combining of file segments between users. (The BitTorrent algorithm cleverly prevents parasite users by matching download and upload rates — eg. you can get 10kB/s of downloads only if other users can upload from you at 10kB/s.)

ABC is happy because they can track download numbers and will have far lower bandwidth bills since users simultaneously downloading the same file will mainly obtain the file from each other rather than the ABC.

Users are happy because there is always at least one copy of the file available — the ABC’s “seed” copy and they are guaranteed that the file they are downloading matches their request.