In .net 2.0, the dispose method of CryptoStream calls the Close method
on the destination stream (verified with .NET reflector). In 1.1, it did
not. In my opinion, the 1.1 behavior was preferable. After all, once I
am done decrypting the data, I may want to look at it. Why should I have
to keep the CryptoStream around for that purpose?

Here is an example. This routine works in .net 1.1, but fails in 2.0.

static public string DecryptString(string s, byte[] key, byte[] iv ,
                               SymmetricAlgorithm encryption_algorithm)
{
    using(MemoryStream msout = new MemoryStream())
    {
        byte[] data = Convert.FromBase64String(s);
        using(CryptoStream crypto = new CryptoStream(msout,
                 encryption_algorithm.CreateDecryptor(key,iv),
                 CryptoStreamMode.Write))
        {
            crypto .Write(data,0,data.Length);
            crypto .FlushFinalBlock();
        } // dispose of the cryptostream, we don't need it anymore

        msout.Position = 0;
        data = new byte[msout.Length];
        msout.Read(data, 0, data.Length);

        return System.Text.Encoding.UTF8.GetString(data,0, data.Length);
    }
}

In 2.0, the memorystream is closed when the cryptostream is disposed by
the inner using statement. It is straightforward to refactor this so the
cryptostream lives as long as the memory stream, but that seems
fundamentally wrong from a resource management standpoint.

H^2