On Wed, Jul 28, 2004 at 03:54:26PM -0400, Hubert Chan wrote: > Adam> it does not come with multiple sub-puzzle support which could be > Adam> added. > > Will this involve another format change? Or should it be backwards > compatible? (I assume that it will change the format.) It would change the format. It is a bit of a tie - PRO gives apps which are about variance opportunity to make it quite low (eg if we defined max 16 sub-hashes) - MINOR CON challenge is a bit bigger when you do it - CON more complex code to create - CON more complex code to verify > Adam> The other part missing from the implementation is the hooks to > Adam> make use of extensions. ... > > OK. I don't think it should be that much of an issue if the Debian > package lacks some feature. I just don't want Debian users to be > generating incompatible stamps for the lifetime of the next stable > release (which some Debian developers tell me should be much shorter > than the current stable's lifetime, but I'd rather be cautious). Yes I was thinking the extension _framework_ with no implemented extensions should not be so hard to implement. It is just checking for the existance of the hashcash extension by-name in the search-path, if its there exec it if not ignore it. So just some system / fork/exec stuff. If someone has some simple, readable and secure BSD / PD licensed code to do this kind of execing and look at status code handy. > Adam> Plus need to integrate Chromatix fast hashcash code. Think main > Adam> issue there is to have a way to know which code to run on a > Adam> machine based on eg some CPU properties code so it is fast, or a > Adam> very short trial of the set of alternatives that run on the CPU to > Adam> measure which is best. > > I guess the best thing to do is to measure once (e.g. at package install > time), and store the result for which alternative is the best, right? Ick. I was hoping either the test was fast enough to run always, and the CPU detection was also v. fast. > Adam> And the last and probably biggest issue is how to deal with CPU > Adam> inflation. I was thinking this could be implemented using the > Adam> extension mechanism. One way is to pass around p2p the largest > Adam> required-hash definition seen so far, and have someone / some > Adam> group changing this over time using some pre-agreed criteria. > > OK. I guess this is another feature that is less of an issue for the > Debian package to be missing, since it can always be done by hand. That's a quite good pragmatic suggestion. With the above, plus I think Chromatix suggestion to use clock() vs wall-clock time (which is not good on systems with load), and I think this ought to be doable quite quickly. Adam