multi-homing fixes

> Do you *really* want your DNS TTL set down in the same range as
> the time for a BGP route fall-over?

Ever read RFC1123?

It states:
2.3 Applications on Multihomed hosts

      When the remote host is multihomed, the name-to-address
      translation will return a list of alternative IP addresses. As
      specified in Section, this list should be in order of
      decreasing preference. Application protocol implementations
      SHOULD be prepared to try multiple addresses from the list until
      success is obtained. More specific requirements for SMTP are
      given in Section 5.3.4.

      When the local host is multihomed, a UDP-based request/response
      application SHOULD send the response with an IP source address
      that is the same as the specific destination address of the UDP
      request datagram. The "specific destination address" is defined
      in the "IP Addressing" section of the companion RFC [INTRO:1].

      Similarly, a server application that opens multiple TCP
      connections to the same client SHOULD use the same local IP
      address for all.

Unfortunately, many programs have chosen not to do this.

Well, yes.

But I thought the thread was talking about multi-homed _networks_,
which I take to be a different problem from multi-homed _hosts_.

Am I in over my head again?

If we define a network as something, which consists of: clients and
servers (hosts) - then the problem on hosts could be solved with DNS as
Clients could be either:
1. NATed behind a HSRPish device so that they always use the same IP,
   and the NAT device translates to whichever uplink is currently active
2. Reconfigured if one link should be down