So I was wondering if there are any major differences between the various implementations of the hash algorithms, take the SHA series of algorithms for example. All of them have 3 implementations each, 1 in managed code and 2 wrappers around different native crypto APIs, but are there any major differences between using any of them? I can imagine that the wrapper versions could have higher performance since its being executed in native code, but surley hey all need to perform the exact same calculations and thus provide the same output ie hey are interchangable. Is this correct?
For instance SHA512CNG cant be used on XP SP2 (docs are wrong) but SHA512MANAGED can.
@Maxim – Thank you, but not quite what I was asking for. I was asking if there is any difference, other than possibly performance, from using the Managed/CryptoServiceProvider/CNG implementations of a given hash algorithm. With .NET 3.5 you get all of the hash algorithms with three implementations, so
SHA512Managed SHA512CryptoServiceProvider SHA512Cng
The latter two being wrappers around native APIs. This is true for all SHAxxx implementations for example.
One difference is that the native versions (at least some of them) are FIPS-certified (i.e., approved by the US government), whereas the managed ones are not. If your code happens to be running on a Windows machine that has been configured as ‘FIPS only’, attempts to use the managed versions will fail.
Most Windows machines are not configured in that way, but if you’re deploying to a government- or defense-oriented (or other highly secure) environment you may run into this situation.
See http://blogs.msdn.com/shawnfa/archive/2005/05/16/417975.aspx.