On Fri, Jul 2, 2010 at 2:55 PM, RJ Atkinson <rja.li...@gmail.com> wrote: > > However, it is important first to note that ILNP *by itself* > enables a multi-path TCP implementation -- without requiring > any TCP protocol extensions/options. Further, ILNP's approach > also enables any other transport-layer protocol to instantly > become multi-path capable.
Interesting! Could you explain in more details or point me to a paper how this is achieved - couldn't find it described in http://tools.ietf.org/html/draft-rja-ilnp-intro-03, thanks! > > To provide a single example of how ILNP can enhance existing > proposals for MP/TCP, combining ILNP with MP/TCP enables one > to use a single IPsec session with a multi-path TCP connection. > By contrast, with current IPsec, one would need to have a > separate IPsec session for each pair of TCP endpoints. So > the key management complexity associated with MP/TCP > would be reduced from O(M * N) to O(1) -- which is a very > significant benefit and quite complementary to existing work > within the MP/TCP area. True, since the IPsec session is "anchored" to the identifier and thus agnostic to transport protocol changes. But IPsec is not much used today, SSL/TLS has become the major player except in site-to-site VPN solutions. tcpcrypt is an interesting approach - but it is only for TCP -- patte _______________________________________________ rrg mailing list rrg@irtf.org http://www.irtf.org/mailman/listinfo/rrg