Re: Next question...

[Brian E Carpenter <brian@hursley.ibm.com>]

I'd say yes, for some definition of policy we need policy control,
and it needs to be global rather than per-host.

OTOH I wouldn't exclude that the actual *enforcement* of the policy
happens in the hosts, under a site-wide policy broadcast in some
way.

  Brian

Tony Li wrote:
> 
> |   > Have we reached consensus that we need to deal with multihoming
> |   > policies?
> |
> |   I am sorry but i do not fully understand what do you mean by this.
> 
> As soon as a site is multihomed, there are multiple possible ways
> of forwarding packets, both inbound and outbound.  Whenever a human
> is involved and there is a choice to be made, we end up giving the
> human a say in how that choice is made.  Such dictums we generally
> refer to as 'policy'.
> 
> In this case, we would like to be able for some human to tell us
> which packets should exit on which link and which packets should
> arrive on which link.
> 
> Certain policies are unimplementable even tho they are desired.
> I don't mean to open the discussion about which policies we will
> support, but I was hoping that we agreed that we need to provide
> SOME policy controls as part of our solution.
> 
> Tony