70 likes | 164 Views
Issues in the SEND base document draft-ietf-send-ndopt-00.txt http://www.piuha.net/~jarkko/publications/send/issues. 58th IETF, Minneapolis Jari Arkko, Ericsson Research NomadicLab Pekka Nikander, Ericsson Research NomadicLab. Issue Statistics.
E N D
Issues in the SEND base documentdraft-ietf-send-ndopt-00.txthttp://www.piuha.net/~jarkko/publications/send/issues 58th IETF, Minneapolis Jari Arkko, Ericsson Research NomadicLab Pekka Nikander, Ericsson Research NomadicLab
Issue Statistics • Between draft-arkko-send-ndopt-00 and draft-ietf-send-ndopt-00: • 25 issues • 3-4 editorial, the rest were technical • many technical issues solved when ND options chosen over IPsec • Currently 6 issues not fixed in the document • 5 with a proposed solution • 1 open one http://www.piuha.net/~jarkko/publications/send/issues
Issues Resolved, not in the latest draft: • 13 - Editorial issues from Pasi and Valtteri • 30 - Change back the RSA-PKCS version number Being discussed: • 27 - DCS and DCA semantics • 28 - DCS source address • 31 - Source address selection and CGA addresses • 17 - Different functions of Redirect
27 - DCS and DCA Semantics • DCS and DCA transmission rules largely copied from RS and RA • It would be more natural to just trigger the process from receiving a message with an unknown signer • The two additional things that are required: • Rate limitation • Allow the router to unicast or multicast the message, depending on situation • As a result, periodic DCA is not needed • Text proposal worked out on the list, appears to be adopted
28 - DCS source address • Background part 1: DCA source required to be link-local • Background part 2: RA source L-L, RS source anything • Current specification: DCA source is anything • Question: Should we restrict this to saying its link-local? • What if someone sends a DCS using a wrong global address, say, from a nearby link? • (I may not fully understand why 2461 did things in the way it did) • Seems safest to require DCA source to be link-local?
31 - Source Address Selection and SEND • Is there a dependency between source address selection and SEND? • If there are CGA and non-CGA addresses, SEND can only provide security if CGA addresses are chosen • Should there be a rule to prefer CGA addresses? • What the priority of the rule should be? • Should there be an API to switch the default? • Suggestion: • Generally, this does not matter as CGA is normally on/off • There should be a rule, in one of the SEND documents • The priority of the rule should be high • No API is needed, RFC 3484 already has a config table
17 - Different Functions of Redirect • The design of SEND only considered Redirect as a message to inform of a better router • Francis pointed out there are other functions: • Link-layer address update by the node itself • Specify link-layer address of an off-link destination which is in fact on link (ATM & NHRP) • We can easily take care of the editorial part of this issue • Any security implications? • Redirect from a host would use CGA, from a router certs • Appears to be OK, as long as the document takes care of this