Re: draft-macaulay-6man-packet-stain-00 --- Issues to consider

[Ehud Doron <[e-mail]>]

>
>  Hello Tyson
>
>
>
> Indeed a very good and interesting work!!
>
>
>
> I would like to suggest a few issues for you and the WG to consider:
>
>
>
> 1.       In chapter 5 you proposed the PSDO structure. I think you must
> consider the addition of Stain structure and semantics to be part of the
> draft. Make the Stain semantics be vendor or security managers dependent
> can be problematic. We must consider enterprise connecting to multiple ISPs
> and vice versa. For Stain provider, define a Stain per customer might be
> difficult to manage. The Stain provider cannot impose Stain structure
> because his customers might be connected to multiple ISPs, thus to multiple
> Stain provider.
>
> 2.       Page 4, chapter 3, item #3 :  I think that major concern here is
> *scalability*. The client must deal with huge number of security
> intelligence, this imposes relatively complicated and expensive hardware
> and software. In chapter 3.1, it might be reasonable to add the scalability
> benefit. In system view and in general, the IP Staining approach introduces
> a genuine and efficient architecture with relatively *small number* of
> high scale devices (PMDs) and *large number* of simple and low scale
> devices all over the network, mainly in costumers networks. I think this
> can be emphasized also in chapter 3, and helps to better understand the
> huge benefit of the proposed architecture.
>
>
>
> 3.       Page 8 – The S and U bits shall be better explained. There
> usages are not clear enough.  Also please consider to add a requirement in
> the Option Type to allow the PSDO to be silently ignored by any nodes don’t
> supporting staining, specified by the 2 high order bits in the Option Type .
>
>
>
> Best wishes,
>
>
>
>
>
> Ehud Doron
>
> Senior Architect, CTO office
>
> Radware Ltd.
>
> www.radware.com
>
> T:[phone]
>
> M:[phone]
>
> F:[phone]
>