[doap-interest] New DOAP repository: doapspace.org

[Stuart Yeates]

In message <11BCE37B-768A-4581-8CA5-89445A8DF3D9 at bblfish.net> Story Henry <henry.story at bblfish.net> writes:
> I have a little java applet for creating sha1sums:
> http://blogs.sun.com/bblfish/entry/my_bloomin_friends
> 
> Let me know if I got something wrong!

Bloom filters are an interesting idea, but I think I see a problem. 

For them to work, they need to have a reasonably low rate of false positives and the rate of false positives fro bloom filters is related not only to the number of emails in the hash, but also to the number of emails the hash can be tested against. 

We could easily imagine global social network with a billion individuals, each of whom have ten email addresses in use.  Some of those people are going to be natural networkers and build up thousands of contacts over their lifetime, maybe tens of thousands. It's going to take a long hash to avoid creating a loose network of erroneous connections between those people.

Even worse, there is currently no mechanism (that I know of) for asserting the falsehood of RDF triples once errors start circulating.

cheers
stuart