Re: [freehaven-dev] OceanStore

[dmolnar@belegost.mit.edu]

On Sat, 6 Jan 2001, David Hopwood wrote:

> This strikes me as extremely ambitious - so much so that I doubt it will
> be fully implemented in the near future. A useful subset of it might be
> implemented, though.

Yes - probably the same thing will happen to Intermemory. What bothers me
about OceanStore is that the current "People" page lists only one person,
an undergraduate, as responsible for security. I'm sure he's an extremely
capable person, but still.


> 
> I don't think that trying to do any useful manipulations on encrypted data
> (section 4.4.2) will work in practice; appending an update log and then
> occasionally consolidating and re-encrypting the object should work, though.

For what it's worth, the paper on "Search on encrypted data" is at
http://paris.cs.berkeley.edu/%7Eperrig/projects/se/se.ps

I haven't read it yet. 


> 
> Note the *big* assumption that faults are uncorrelated in the analysis
> in section 4.5. I'm not sure this is reasonable even in the absence of
> deliberate attacks.

Yeah...the Tornado code analysis only goes through if the faults
are uncorrelated. This came up last spring when we were considering
Tornado codes for Free Haven.

-David