Re: [Cfrg] Asking comments: New Version Notification for draft-kiyomoto-kcipher2-00

Hello Joachim,

> This is where soo many cipher designers go wrong. From an implementer
> point of view having access to test vectors and a simple, functional
> reference model which matches the algorithm description is almost
> essential. Unfortunately test vectors is rare. And a good, reference
> model is rarer still.
>
> What you often get is an implementation which is unrolled or in other
> ways reorganized for maximum speed, have variable names not matching the
> written algorithm description, lacks comments is generally written in
> badly coded C.
>
> Yes, a version optimized for speed is good too, but that comes after
> having the functionality down. And using single letter variable names,
> bad intendation and lack of comments does not lead to faster
> implementations.
>
> So, please if possible, include the simplest, cleanest, well coded, well
> documented albeit slowest implementation that is self contained (besides
> stdint, stdio etc) do as little macros, typedefs etc as possible. And
> compiles easily.
>

I really appreciate your comments. Due to my lack of experience in the
standardization process and the field of cryptography, I think I
misunderstood the purpose of the standardization and the use of test
vectors. Now I can see what Jim and Jon were looking for.

I will revise the draft and rewrite the codes, so that the procedures
in the implementation can be shown clearly.

Thank you again,

Wook

On Fri, 08 Apr 2011 15:39:02 +0900, Joachim Strömbergson  
<Joachim@strombergson.com> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Aloha!
>
> On 2011:04:08 6:49, Wook Shin wrote:
>> I just wished the source would be like an extra appendix which might
>> provide some helpful insight for someone when she/he reads the draft.
>> I don't think the source code is the unique and the most efficient
>> implementation of KCipher-2, though...
>
> This is where soo many cipher designers go wrong. From an implementer
> point of view having access to test vectors and a simple, functional
> reference model which matches the algorithm description is almost
> essential. Unfortunately test vectors is rare. And a good, reference
> model is rarer still.
>
> What you often get is an implementation which is unrolled or in other
> ways reorganized for maximum speed, have variable names not matching the
> written algorithm description, lacks comments is generally written in
> badly coded C.
>
> Yes, a version optimized for speed is good too, but that comes after
> having the functionality down. And using single letter variable names,
> bad intendation and lack of comments does not lead to faster
> implementations.
>
> So, please if possible, include the simplest, cleanest, well coded, well
> documented albeit slowest implementation that is self contained (besides
> stdint, stdio etc) do as little macros, typedefs etc as possible. And
> compiles easily.
>
> (Note: This is not a critique of kcipher2 but a general observation.)
> - --
> Med vänlig hälsning, Yours
>
> Joachim Strömbergson - Alltid i harmonisk svängning.
> ========================================================================
> Kryptoblog - IT-säkerhet på svenska
> http://www.strombergson.com/kryptoblog
> ========================================================================
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.11 (Darwin)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
>
> iEYEARECAAYFAk2erYYACgkQZoPr8HT30QH+4ACePO0asRvreR9t5QXUVKRe53Bv
> 7P8An3/hTwwtsNqmeOxFifLLFTzR6Wte
> =yjke
> -----END PGP SIGNATURE-----
> _______________________________________________
> Cfrg mailing list
> Cfrg@irtf.org
> http://www.irtf.org/mailman/listinfo/cfrg

-- 
Wook Shin
Information Security Laboratory,
KDDI R&D Laboratories, Inc.