IPng Implementors Meeting DEC Cambridge Research Labs Kendal Square, Cambrdge MA Oct 10, 11, 12 1994 ------------------------------------------------------------------------ Oct 10 Topics Auto-Addresing Transition Security Mobility ICMP Misc ----- ICMP Make packet too big it's own message type? Needs parameter to specify MTU. AGREED Multicast Fragmentation? Allow Pkt too big message to be sent to source. Deering wants MTU discovery on multicast links. Others want to have routers do fragmentation for multicast traffic. Allow pkt to big in response to multicast. No router fragmentaton. How much payload to send in response in ICMP error messages. Fill up to 576 or all of packet. Keep 576 limit. Ptr could point off end, spec needs to make this very clear. Take contents out of redirect. Wasn't thouht as needed. Changed mind, keep current behavior. ------- Auto Configuration (dave katz> Overlap in functionality between auto config (almost draft) and discovery drafts. Focus is to take address config out of DHCP. Functionality goals/constraints: dentists office, campus, low administrative mode, high administrative mode, single host operation. Current scheme is request/response. Not broadcast by router, host listens. Requires some sort of agent. First step is host creates local use address. Prefix +-----------+--------------+--------+ | 1111 1110 | 0000....0000 | Cookie | +-----------+--------------+--------+ Discussion on need for address type before cookie. Should remove it. Next step is Issues: More than one response? Lease time Determinism Change time field to 32bits, from 16bits. -------------- Transition -------------- Text Representation of Addresses Now 1:2:3:4:5:xxx Proposed: ?????? Stick with current formats. ---------------------------------------- Wednesday 10/12/94 Security Scott: Talked about security requirements. Suggest sending comments in response to last call if unhappy with security "mandatory" requirements. -------- Mobility Two proposals: Charlie Perkins and (vs.) Bill Simpson Perkins: End-to-End option type Send "binding === Correspondent Host has bindings CH ===========> FA ----------> MH Mobile Host is caching ..... Deering suggested having a mobile host acquire a tempary address on local subnet instead of using a foreigh agent. Use source route to distribute bindings to Notes for Ross: Simpson discussed his mobile ip scheme. Deering asked questions about how long the bindings will last. Mobile host and foreign agent might be same machine. Might violate IBM patent. Question about mobile routers. Simpson says covered in his proposal. Simpson thinks there is a concensus on these mechanisms. Perkins thinks desirable to reduce xxxx between mobile host and agents. Deering thinks we should just do what is done in IPv4. Some things in IPv6 (source route, etc.) can improve it in IPv6. Would like all routers to support mobility. Bound think an approach that reduces number of agents is good. Bound think mobility is important to get done soon and important for IPv6. Discovery mechansims can facalite mobility in IPv6 that can not be done in IPv4. Simpson agrees, that IPv6 will be a higher quality product than IPv4. Simpson also mentioned that IPv6 has authentication which will work very well for mobility. -------------- API / Gilligan ..... --------------------