Talk:Pseudo-Hadamard transform

Excuse me, I suppose that there is one error in the first formula in the article. It should look like


 * $$a' = 2a + b \, \pmod{2^n}\,$$


 * $$b' = a + b\, \pmod{2^n}\,$$

and accordingly


 * $$b = 2b' - a' \, \pmod{2^n}$$


 * $$a = a' - b' \, \pmod{2^n}$$

(moreover it should be so if matrix $$H_1 = \begin{bmatrix} 2 & 1 \\ 1 & 1 \end{bmatrix}$$ is correct)

(Of course I'm not specialist, so I will not change the article)--Gistereziz (talk) 16:29, 9 November 2009 (UTC)

Paper by St Denis is pointless
Please do not include the paper by Tom St. Denis. It contains a lot of incorrect theorems and is not relevant. First is claims that MDS transform require O(N^2) time to compute. This is wrong. Many transforms are computable in O(N log N). Then Theorem 1 claims that $$\beta_2=3$$. This is wrong for binary fields. Note that binary fields and rings modulo $$2^n$$ are the cases that are important for cryptography.

The paper is only published on the iacr eprint server. These papers are not reviewed. 67.84.116.166 19:22, 10 September 2006 (UTC)
 * I know that ePrint is not reviewed, and I don't object to removing it on the above grounds (I haven't read the paper, though). I reverted your change originally simply because your edit summary was that the paper was "pointless". That isn't a rationale for removing it, whereas something like "paper is unpublished and contains numerous errors" would have been. &mdash; Matt Crypto 21:17, 10 September 2006 (UTC)


 * How relevant a paper is seems important to me too. Being able as a reader to find important paper fast is one of the advantages of an encyclopedia over Google and other search engines. 67.84.116.166 23:01, 10 September 2006 (UTC)
 * Sure. On academic topics, Wikipedia articles should be quite similar to a survey paper in the sense that it should survey and integrate results from the published body of research on a topic. If there's a lot of research in a topic, then we have to pick the important papers. &mdash; Matt Crypto 23:22, 10 September 2006 (UTC)

According to me the connection to the Hadamard transform should be mentioned. Or the reason why this transform is better than the Hadamard transform for diffusion purposes. As I am a complete novice on this topic I can just guess that one of the advantages is the invariance under modular arithmetics. R. —Preceding unsigned comment added by 87.165.160.119 (talk) 10:25, 10 March 2008 (UTC)