Le 3 août 2011 à 16:09, Qiong a écrit : > Hi, Remi, > > This is just the question I would like to mention. Please see inline. > > Several contributors have expressed, in the context of stateless solutions, > wishes to add flexibility to static port set assignment. > > But, where "stateless" has to mean "no per-customer state", this wish is > self-contradictory. > > Now: > - Some ISP's want _direct CE-CE paths_ .
> [Qiong] : Agree. This is a good feature for stateless solution. > But there might be no big differences from stateful solutions when deploying > CGN on the edge of network, e.g. BRAS, etc. Depending on various factors, he bottom-line cost may be slightly or highly different. One size may not fit all. > > - For this each CE must know mapping rules for all other CE's. > [Qiong]: Agree, especially when applying the co-existence scenario for > exclusive-mode and shared mode. Given the fact the IPv4 prefixes are not > continuous anymore, there might be up to hundreds/thousands of rules for a > certain area. Thousands of rules seems to me a lot. (I keep doubts that, if CE's support statically shared addresses, keeping thousands of IPv4 prefixes would be needed to support IPv4 via IPv6.) In any case, this can be among factors that differentiate which solution applies best to which network. > > Although I think that it would be not a big problem for stateless GW to > handle these rules, I would still doubt the possibility that a customer-side > CPE dealing with the same amount of IPv4 prefixes with network-side GW. Adapting a CE to up to 1000 rules doesn't seem difficult to me, and with O(log n) matching this can be done with satisfactory performance. (More details would be private consultancy ;-)). > And another question, when should we announce these rules into CEs and BRs. At the same time as other parameters (once in awhile to avoid lifetime expiry). > Since in CE-CE situation, a source should always send a packet with the right > destination port in the valid range of destinations. Is it has to been > guaranteed by individual applications ? Or does it have some kind of > limitations depending on applications ? Applications that don't depend on well-known ports continue to do what they do behind NAT44's. How applications get A+P addresses of peer applications is out of scope. Hope it helps, Regards, RD > > Thanks. > > Best regards > > Qiong Sun > > > - Mapping rules must therefore not depend on any per customer state. > - This doesn't permit dynamic increase of CE-assigned port sets. > > If I missed something, please let me know. > If not, agreement on this point could facilitate the ongoing discussion. > > Thanks, > RD > _______________________________________________ > Softwires mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/softwires >
_______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
