From: Mario Contestabile (marioc_at_[hidden])
Date: 2000-02-20 12:35:21
I've been using Yarrow 0.8.7 for well over a year (under Win9x & NT).
its usage under Windows is satisfactory, the next version of our software
will drop that implementation, and use our own, for the following reasons:
- The counterpane implementation of the Yarrow algorithm contains errors
in its usage of SHA-1. At my company (Zero-Knowledge Systems) I'm fortunate
to work with some very strong cryptographers (Ian Goldberg, Adam Back) who
have noticed this. Adam has written a 'C' API which implements the Yarrow
algorithm in a much simpler way, primarily because:
- The counterpane implementation of Yarrow isn't cleanly decoupled from
the entropy gathering, os specific functionality. Ideally, one would want
the algorithm implemented in a class which provides methods for randomness
input & output. That implementation is divided into 3 DLLs, one called for
the core routines, one for safe memory allocation, & one for hooks. In the
core prng routines, there are things like "__declspec(dllexport)" and
- The need for safe memory allocation/deallocation is overkill.
- And finally , to have more x-platform code. That implementation is too os
specific. That algorithm implementation should be able to stand on its
own regardless what the source of entropy is, or where it executes
(kernel, user land).
>> Another idea (not sure where I heard it first, maybe Schneier also) is
>> to have a rng that the user can churn with randomness from the system.
>> This randomness would get added into the mix.
>Why not just incorporate Schneier's Yarrow implementation, which can be
>downloaded from his site? Admittedly, it's beta at present, but the code
>freely available, and does a pretty good job of what you're talking about.
>I seem to recall some Windows dependencies, but it shouldn't be too hard to
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk