Hi list! Apologies in advance for my broad ignorance; as soon as I make some progress in one field, whole new frontiers come into view :)
I just subscribed a few weeks ago, so I didn't have the pleasure of seeing elcalamartevigila's mail when it came in, but I'm interested in the same "space". I would like to write distributed web applications that use some secure, distributed key-value(s) store. Specifically I am interested in distributed data structures. Tahoe-LAFS seems like a great start, but it seems to currently lack support for some paradigms. * It seems that Tahoe-LAFS is most appropriate for "friendnet" operation, and is not so supportive of larger-scale, more intermittent networks, with more "edge" storage than "center" storage. * Relatedly, Tahoe-LAFS does not seem to provide for caching of "popular" content. * I don't see a good way to support "seed-your-own"-style operation: where you have data that you care about, that you would like to expose to the global DHT, but which you need stronger guarantees of persistence. A friendnet persistent layer combined with some more robust global DHT sounds like it could work. * I don't see a way for someone to securely send a message to you using Tahoe-LAFS. This is related to the directory rollback problem, and could be solved if Tahoe-LAFS exposed all values associated with a key, as is done with Kademlia keywords for example. * However, it does seem that one can profitably build interesting datastructures on top of simple k-v stores, as Tahoe shows, though it might require good programming language support for asynchronicity, multiple returns from lookup procedures, etc. But, I have learned a lot from Tahoe, and expect to keep on learning. Thanks for the software, papers, and really weird mutable file algorithm ;-) Cheers, Andy -- http://wingolog.org/ _______________________________________________ tahoe-dev mailing list [email protected] http://tahoe-lafs.org/cgi-bin/mailman/listinfo/tahoe-dev
