Re: A tunneling proposal

["R.P. Aditya" <aditya@grot.org>]

On Mon, Jul 16, 2001 at 02:40:48PM -0700, Ramakrishna Gummadi wrote:
> It all depends on what the first priority is---if it is scalability, then,
> I think, to make TCP/UDP survive all failures, one needs an approach like
> GxSE, which affects the IP layer. If you are willing to be less than
> perfect, and don't want to change the host stacks, tunnels may be a good
> choice, especially when applications can use SCTP to get the
> required reliability.

one of the main problems with the tunnel approach (and I'm not saying there is
indeed a currently practical or better way) is that it requires a degree of
cooperation between ISP-A and ISP-B, which also means that the "decision" is
made independently of the end-network. It moves the redundancy from a "local"
solution up the AS hierarchy, which is what the *LA design was about. One of
the desireable features of current IPv4 multihoming is the ability to solve
for "local" connectivity loss (local-loop/router failure) via "local"
redundancy.

Unfortunately, apart from GxSE or some other "overlay" address scheme that
embeds "real-world" (read geographical) information, you can't fully
compensate for the loss of routing state information that a path-aggregated
DFZ presents and expect to preserve the desireable IPv4 multihoming
characteristics (cf. jabley's survey).

Adi