2013-01-29 10:50, Ole Troan <[email protected]> : > ... >> (c) Once the rule is looked at, getting the PSID is easy (no more magic >> needed than to get find the PSID length ;-)). > > unless you are in 1:1 mode.
There is a point here which AFAIK has never been discussed before. (A reference is welcome if there is one I missed). This point concerns 1:1 mappings (if needed in MAP-E), when combined with shared IPv4 addresses: - To be assigned a shared address, a CE needs a PSID in addition to its IPv4 address. - Currently, this PSID cannot be found in the 1:1 rule itself (ref. http://tools.ietf.org/html/draft-ietf-softwire-map-dhcp-01.) The approach that is in my understanding implied by Ole is that each concerned CE will be delegated a MAP-E IPv6 prefix, in addition to that used for native IPv6 addressing, with the PSID in it. (The detailed format isn't specified yet AFAIK, but specifying one is possible.) Another approach, AFAIK much simpler in operations, is to add a rule parameter to assign a PSID, present in 1:1 rules. This eliminates the need to delegate to each CE an additional IPv6 prefix. Example: . CE delegated prefix: 2001:db8:0:100::/56 . BMR for this CE Rule IPv6 prefix : 2001:db8:0:100:/56 EA-bits length : 0 Rule IPv4 prefix : 1.1.1.1/32 PSID : 0x11 Conclusion: if 1:1 mappings are part of MAP-E, add a PSID parameter in 1:1 mapping rules. (Otherwise: (1) specify detailed formats of MAP-E specific delegated prefixes > /64; (2) explain why delegated prefixes used for IPV6 aren't found sufficient.) Regards, RD >>>> 2001:db8::1 >>>> Rule IPv6 prefix: 2001:db8::1/128 >>>> Rule IPv4 prefix: 1.1.1.1/32 >>>> EA-bits length: 0 >>> In addition, it has the desirable characteristic of creating an IID. >> >> A constant IID was the logic proposal for the original Encapsulation >> solution (then called "4rd" before becoming "MAP-E", ref. >> tools.ietf.org/html/draft-murakami-softwire-4rd-01). >> It is only when we tried to merge -E it with -T that we abandoned this >> simplicity. > > simplicity for whom? > the interface-id is constant for the MAP node. I have not heard any > implementors state it is hard enough to be worth > arguing over whether to write an IPv4 address in there rather than a set of > zeroes. > what is simplest for the operator? having to calculate the IPv4 address based > on the rules? or look at the IPv4 address in the IID? > >> (Incidentally, to avoid limitation (b) above, we introduced that at that >> time a PSID-length field in IIDs (ref. >> www.ietf.org/mail-archive/web/softwires/current/msg02994.html), but this >> complexity would be superfluous for MAP-E.) > > to summarise, your proposal is to overlap with the subnet router anycast > address, i.e. all zeroes? This has been discussed long ago. Following a comment made by Wojciech if I remember right, the constant value can be a 1 to avoid any risk of conflict with the subnet router anycast address. > > cheers, > Ole _______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
