User privacy is very important to BitTorrent, Inc. I am sure all client implementors are cognizant of their many competitors. Fearing exodus, most implementors probably do not stray beyond reporting traditional tracker statistics and contacting auto-update servers.
In particular BitTorrent mainline and uTorrent clients report NO statistics about how clients are used except tracker stats, and these tracker stats are not reported to BitTorrent, inc., but rather to the tracker(s) for each torrent the client is downloading or seeding. Thus BitTorrent, inc. has less information about how its clients are used than do many public tracker sites. BitTorrent, inc. logs every contact to our autoupdate servers, and we have tracker logs when someone downloads from BitTorrent.com.
Autoupdate logs give us some notion of how many people use BitTorrent mainline and uTorrent, what versions they are using, and how many days out of the year and at what time in the day they turn on their clients, but autoupdate logs provide no further insight into how the clients are used. BitTorrent does not have statistics on how many people use BitTorrent mainline or utorrent and have autoupdate turned off.
Since the vast bulk of downloads are not from BitTorrent.com and because BitTorrent.com swarms are much smaller than many swarms found on popular public tracker sites, our tracker logs provide a very limited peek into how clients are used.
We are unable to provide good answers to some very basic questions, such as:
- how many swarms does a user typically join?
- how many files does a user queue for downloading?
- how much peer capacity goes unused?
- how well do clients allocate capacity to swarms that really need capacity?
- how many trackers does a client typically talk to simultaneously?
- what are the seed/partial seed/downloader distributions on each of the swarms a client has joined or has queued?
Answers to these questions would enable researchers to build models of user behavior which could be used to optimize clients to better allocate resources. Furthermore such questions can be answered without revealing user identity, and without revealing what is being downloaded or from whom.
BitTorrent, Inc. will not gather statistics to answer these questions; therefore, it falls upon the community to willingly aid in the research and development of the BitTorrent protocols.
Anyone willing to write an instrumented variant of an open-source client and release it via bittorrent.org, please contact the editors of BitTorrent.org or post to the forums.
Are users willing to participate by using a client that reports anonymous statistics?
Any gathered statistics would be made publicly available via bittorrent.org.
If considerable number of users are willing to use such a client we can develop (or help to develop) something useful for the research community....
I guess number of users should be several 10's of thousands as the network include millions of users, otherwise it would not be representative of the actual network