How do I use it?

You can access the data by constructing a URL:

http://tfeed.maxwell.syr.edu:8080/proc/peers/GetPeers/<path entry>

working examples:

http://tfeed.maxwell.syr.edu:8080/proc/peers/GetPeers/news.maxwell.syr.edu
http://tfeed.maxwell.syr.edu:8080/proc/peers/GetPeers/newsfeed.cwix.com

Once you've navigated in, you can follow links to incoming or outgoing peers.

What are the numbers after the host name in the display?

Each entry has the format:

Why is it slow to do lookups?

Hard to say. The web server on port 8080 is a single process (one thread), and must talk to the mapping software (one request at a time) to satisfy each request. If more than two or three people are simultaneously doing lookups, it will slow down. If you're having consistent problems with slow lookups, please send email to news@maxwell.syr.edu

Why do I see entries with a path entry "(null)"?

This is because the list is of known path entries is only read on interface startup. It is too expensive to load them for each request (there are currently 89000+ entries) and I haven't yet written code to reload them at an interval.

How does it work?

The path mapper takes a feed (all articles < 10k) from a Maxwell Usenet news server. It scans the article for its path line and discards the remainder (noting the size of the article). The path line is then processed in pairs starting from the right and moving to the left. For each pair, the left host gets the right marked as an incoming peer, the peer's count incremented by one, and the peer's byte count incremented. The right host gets the left marked as an outgoing peer, its count incremented by one, and its byte count incremented. Both also get the path length added in to allow computation of an average.

How often is it updated?

It is updated in real time. If look at news.maxwell.syr.edu in your browser, wait 5 seconds, and hit "Reload", the counters should change. If not, the feed's not running, or there's some other failure.

The statistics are cumulative from when the database was initialized. The database may be dumped occasionally as software is updated or changed.

How accurate is it?

Unknown. The parsing code was not written to cover all possible scenarios, it is relatively simple and designed to be accurate only in gross accounting. It is important to note that you only see the view of the world as it appears from our feed. Some active hosts may not even appear since we never see any articles that pass through them.

What platform does it run on?

The parsing is taking about 15% of a 266Mhz Digital Alpha running FreeBSD.

Is the code available?

Yes. download

Is there any future for this?

Unknown. It would be nice to have a way to coordinate among several sites running the software since it would give a "truer" view of how articles really propagate.

Who wrote it?

Chris Sedore - cmsedore@maxwell.syr.edu