Re: [IPsec] Some comments to the draft-sheffer-ipsecme-pake-criteria-00.txt

[Yaron Sheffer <[e-mail]>]

Hi Tero,

thanks for your comments. Please note that some of them have already 
been addressed in -01, which at the moment is only available on 
http://trac.tools.ietf.org/wg/ipsecme/trac/wiki/TempDocs.

While not disputing the importance of the engineering criteria, I think 
there's quite a bit of differentiation between the various options on 
security and IPR.

Thanks,
	Yaron

On 22.3.2010 0:20, Tero Kivinen wrote:
> The document should list more of the engineering considerations too in
> addition to the IPR and security. The Security requirement is
> mandatory (how that is show is completely different matter). The IPR
> issue would be nice to have. So I expect we end up looking other
> properties too.
>
> The other considerations section is quite short, and for example I do
> not really care whether the protocol was specified in standards
> document or scientific paper, as in any case we are going to make sure
> our document describes the protocol in a way that alows
> interoperability. Also in some cases the standards document are WAY
> too hard to read to be able to provide interoperability... :-)
>
> Also the second point of having protocol already integrated into IKE
> does not really matter, as the only thing it does is to save us some
> time. If the protocol is good and suitable to be integrated to IKE, I
> do not really think we should leave it out just because someone has
> not yet done the work to actually write the document telling how they
> are integrated. This is the document we are going to be writing here
> anyways. More suitable criteria would be to say it must be something
> that can be easily integrated to the IKE.
>
> I think this section should also include more concrete engineering
> criteria lists, like:
>
>    - Number of round trips
>    - Cost of cryptographic processing
>    - Does it share same cryptograhic blocks already used in IKE (I.e.
>      it would be better to have protocol using MODP+SHA1, than some
>      polynominal shares plush some hash function not used in IKE).
>    - Does it share for example the Diffie-Hellman groups already used
>      in the IKE.
>    - Does it fit to the IKE PSK model (i.e for example it should not
>      rely on trusted third party)
>
> Most likely we are going to be using those engineering criteria to
> select the final protocol, as I do not expect there be that much
> difference (that we can agree on) on other security and IPR.