--- Log opened Sat Mar 23 00:00:02 2013 18:43 < gmaxwell> petertodd: it seems to me that all this TPM everything (including hal's stuff) could all be converged on a single computational oracle model. 18:44 < gmaxwell> E.g. you write a TPM-program that takes a AST-program hashroot. And derrives a program specific secret value H(AST-root||oracle_secret) and pushes that on its stack along with the time. .. and runs whatever program the user sends it. 18:45 < gmaxwell> From that single TPM enviroment you could do anything you'd want to have a tpm do. 18:45 < gmaxwell> Seems better than inventing a new TPM program for every usecase. 18:46 < gmaxwell> For reasons of efficiency you'd want varrious cryptographic operators available as instructions, but they could be generic ones. 18:49 < gmaxwell> Arguably TPM is dumb and should have just invented that in the first place. :P 20:34 < petertodd> gmaxwell: That approach makes a lot of sense to me, and not just technically. If you're creating abstract oracles, you can also safely sell hardware implementing these oracles publicly as they are general purpose and can be used for anything. 20:37 < gmaxwell> ah, so even if some oracle usages are prosecuted .. interesting. 20:38 < gmaxwell> I think the AST stuff adds a lot to the oracle, as it even prevents the oracle from knowing the complete program that it participates in, and also compresses large oracle programs. 20:39 < petertodd> Ah, the preventing full AST knowledge is a godo point there too. 20:39 < petertodd> Which in turn means there can be 1 to n oracles actually doing this stuff. 20:40 < petertodd> The crazy thing about this model, is n could actually be really small, and it'd still work, or really large, and it'd still work. --- Log closed Sun Mar 24 00:00:04 2013