https://bugzilla.wikimedia.org/show_bug.cgi?id=28419
--- Comment #64 from Tyler Romeo <[email protected]> 2012-08-06 11:07:14 UTC --- (In reply to comment #61) > I suggest we stop the cryptoparanoia competition here and finally get at least > SHA-2 or WHIRLPOOL (since you people make such a great deal of implementing > PBKDF2 so you can't get it done without tearing each other apart). We're still > using MD5 right now for password storage. > > I could go ahead and patch the system myself. But it looks that it has been > done at least twice, can we finally get *something* into gerrit? Preferrably > in > the small portions. If I were you, I'd start with OOP rewrite of the current > password system without any new backends, then commit a patch with PDKBF2 > backend, etc. There already was something in Gerrit. I submitted a patch earlier last week that fully implemented an OOP password system. You can see the patch here: https://gerrit.wikimedia.org/r/16049 Unfortunately, I abandoned it due to the insane arguments we've been having in this thread. If you want I can adjust the patch (so that it uses hard-coded types like Daniel has proposed) and re-submit it to Gerrit. (In reply to comment #62) > (In reply to comment #60) > > Except it would be preferable to use scrypt over PBKDF2, so we should look > > to > > try and find a platform-independent solution for implementing scrypt in MW. > > Have you tried reading scrypt? > > Firstly scrypt actually uses pbkdf2 inside of it's algorithm. So even if we > did > try writing a scrypt implementation in php the fact that we have hash_pbkdf2 > coming out in 5.5.0 is still a good thing. > > However, after that I looked deeper into scrypt. One of the things scrypt uses > in it's algorithm is an algorithm they call smix which uses a combination of > the ROMix algorithm, BlockMix algorithm, and they use Salsa20/8 for the hash > algorithm. > > Now, if the person trying to implement scrypt in php at this point hasn't > already burnt out from the mere thought of having to implement all these > algorithms in php by themselves. I'll have to point something else out. The > salsa20 hash algorithm was removed from php in 5.4. > > With all that on the table I have a feeling that it is impossible to write a > properly efficient platform-independent version of scrypt in php. Certainly > not > something we're going to write. > > So while we can hope the unofficial php-scrypt module is correctly written and > people can decide to install that and write a MW implementation that uses it. > We cannot use scrypt as a default until someone accepts that module into php > and starts shipping it. Yes I have read scrypt. And it is no small task indeed. I should probably rephrase. I just meant to say that we should try and plan for some support for scrypt, as it is preferred over PBKDF2. The unofficial PHP extension for it is actually just a wrapper for the original scrypt binary published by the author, so we can place trust in its accuracy. Now that I think about it, maybe this is the perfect opportunity to make an extension that uses the new pluggable password system. Once the system is implemented, I'll put together an extension that adds scrypt to the password types. -- Configure bugmail: https://bugzilla.wikimedia.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug. _______________________________________________ Wikibugs-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
