hi Woj, prof. Bao and others, as i discussed with Remi in the thread http://www.ietf.org/mail-archive/web/softwires/current/msg03641.html , i found that RFC6145 leaving DCCP support as optional may cause a asymmetric-process when it is applied for double translation, and there was not an explicit statement on the support for other L4 protocols. i do suggest to consider this issue in MAP-T. my humble suggestion is adding two paragraph in the introduction, after the first paragraph, stating that:
Regarding the L4 checksum calculation at the address-family boundary (CE or BR), RFC6145 mandates TCP and UDP but leaves DCCP as optional. In a double-translation environment, if only IPv4-to-IPv6 translator adjusts the checksum of DCCP but the IPv6-to-IPv4 side doesn't, the IPv4 destination may receive DCCP packets with wrong checksum. MAP-T ensures DCCP support by updating its checksum at CE or BR whenever the DCCP packets are passing through. Listing all possible transportation protocols is not feasible nor necessary. For a transportation protocol not mentioned explicitly here or in RFC6145, a MAP-T-compliant equipment MUST either discard packets of that protocol (as unsupported) or have the corresponding code for the L4 checksum adjustification. * Throughout this document, whenever RFC6145 is cited, the above considerations are also involved, even without explicit statement. if anyone thinks support of DCCP and/or other less-well-known transportation protocols is fairly not significant, please also respond. thanks a lot for the concern. regards, maoke
_______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
