On Wed, Jun 1, 2011 at 11:52 AM, Greg Troxel <[email protected]> wrote: > > Yes, they could. What you are missing is > > 1) that this is a capability system, not an ACL system.
This is true, but I want to caution that what we've designed so far is based on what we think is safe and useful, rather than being limited to a specific dogma. In other words, if someone comes up with a new access control scheme for Tahoe-LAFS which is safe and which provides something that real users want, then we'll definitely implement that, regardless of whether it is called "a capability system" or "an ACL system" (or something else). I'm aware that a couple of security and cryptography grad students have been investigating this sort of thing, so I look forward to what they come up with. > 2) if you handed them a decryption key for normal data, they could grab > and download the data. They could then hand it out. > > Basically, if you don't trust people to keep things secret, you can't > share with them. This isn't about tahoe, or rather tahoe has no magic > bullet for this. Right. Most (but not all) of the most onerous of Tahoe-LAFS's current limitations are fundamental limitations of any distributed system (without effective DRM). If anybody can show me a distributed, DRM-free system which overcomes these limitations then that would be AWESOME because we could then immediately clone their solution for future generations of Tahoe-LAFS. :-) Regards, Zooko _______________________________________________ tahoe-dev mailing list [email protected] http://tahoe-lafs.org/cgi-bin/mailman/listinfo/tahoe-dev
