Scope of SA change in design document..
Tero Kivinen <[email protected]> Fri, 11 Nov 2005 20:16:29 +0200
| Newsgroups | gmane.ietf.mobike |
|---|---|
| Message-ID | <[email protected]> |
Mohan Parthasarathy writes: > I think there is a bigger issue here. You can't have > one SA and have multiple flows/packets going over > different paths because if their path charecterestics > are different, the replay check would drop the > packets. Unless ESP can maintain per-address > sequence numbers, how this would be possible. > Perhaps this could be mentioned as a design point. > I don't remember whether this was discussed before > or not. Actually that is not problem with issue 8. The issue 8 is about moving only part of SAs created between two hosts. One SA always will use exactly one address pair regardless of issue 8. Our charter do rule out load balancing, i.e. using multiple addresses at the same time for single IPsec SA is explicitly ruled out for the group. We do already offer a way for the host to create separate group of SAs which are moved separately from each other, i.e. applications can create multiple IKE SAs, and create the IPsec SAs to the IKE SA having similar IPsec SAs. I.e. if you have 3 different groups of IPsec SAs wanting to have different charasteristics from the interfaces uses, you can create 3 IKE SAs and create those IPsec SAs inside each of those IKE SAs. Then you have 3 groups (IKE SAs) which you can separately move from one address pair to another. You can even move the IPsec SAs from one group to another, by recreating the IPsec SA in diffenet IKE SA and then deleting the old IPsec SA from old IKE SA. The problem with ESP sequence numbers comes only if we try to do load-balancing, i.e. use multiple different ip-addresses for the SAME IPec SA at the same time. -- [email protected]