Re: [karp] Handling of LDP hello packets and the crypto table

[Acee Lindem <[e-mail]>]

This would be fine with me. Note that since LDP discovery (UDP) can be
1-to-many and LDP sessions are 1-1 (TCP), the deployment model may require
different key-table interface selectors anyway. However, the change you
propose would allow a deployment model where one key is used for both.
Thanks,
Acee 

On 9/18/13 5:02 PM, "Sam Hartman" <[e-mail]> wrote:

>>>>>> "Acee" == Acee Lindem <[e-mail]> writes:
>
>    Acee> Hi Sam, I don't necessarily agree that this one use case
>    Acee> should greatly complicate the proposed key table entry
>    Acee> format. However, I'm willing to listen to your proposal on how
>    Acee> to handle this.
>
>I don't think it should change the format at all.
>
>Some document needs to describe how to use key tables for securing LDP.
>That document needs to describe what algorithms should be used.
>
>My proposal is that document has joint algorithms that map both to a
>TCP-AO algorithm and  an UDP protection algorithm.
>
>so you have a table in the document with three columns:
>
>1) key table algorithm
>2) TCP-AO algorithm to use when this key table entry is used
>3) UDP protection algorithm to used when this key table algorithm is
>used.
>
>The alternative is to register two key table protocols (LDP-TCP and
>LDP-UDP) with separate algorithm lists and require the admin to enter
>two rows.
>
>I thnik encoding a constant table from the document describing how to
>use key table for LDP into the implementation is relatively easy.
>Probably easier or same level of effort as supporting LDP-TCP and
>LDP-UDP as separate key table protocols.