Hi Satoru, Please see inline.
Best regards! -------------- Yuchi Chen >Hi Yuchi, > >On 2012/11/14, at 5:56, "Yuchi Chen" <[email protected]> wrote: > >> Hi Satoru, >> >> I think the main issue here is not about "naming", but about the original >> motivation of MAP. AFAIK MAP draft promotes a stateless provisioning method >> using mapping with address and port. Although the algorithm MAP draft >> promotes is kind of complicated and couples v4 and v6 addresses, which >> introduces a potential threat of flexibility, > >MAP introduce complication and flixibility threat!? Not at all. >So then, could you answer how much complicated is complicated, and how much >flexibility do you need? > [Yuchi] I don't think the discussion on definitions or degrees of 'complicated' and 'flexibility' is so constructive, and I don't think it will make any sense here. >> but it's well designed and beautiful indeed and makes MAP attractive to >> those who want such a stateless solution. If MAP turns out a stateful >> solution in 1:1 mode (no matter what kind of state it's like), I don't think >> it will be so charming to these people any more, and I think people who >> wants 1:1 may prefer to another solution which is much simpler, satisfies >> the 1:1 scenario well and does not introduce any flexibility problem. > >Why 1:1 is stateful? >I'm not clear where is the threshold to distinguish stateful and stateless? >N=2, 3, or 4??? > [Yuchi] I believe that the differences between stateful and stateless are well discussed in the previous discussions. The point I want to figure out here is that MAP 1:1 mode is not so 'stateless' as other parts of MAP-E. >> >> It will be appreciated if you would so kind clarify the real motivation of >> MAP, and the meaning of 'mapping' you used to describe the DS-Lite (which is >> a well-known stateful solution) if possible, and thus it may help us keep >> discussion more constructively. >> > >Doesn't AFTR maintain mapping of remote IPv6 endpoint address with IPv4 >address and port? > [Yuchi] Actually there is a 'binding' table on the AFTR, which maintains matching between v4 A+P and v6 A, and there's no 'mapping with address and port' in DS-Lite or Lw4over6 as in MAP. The main point here is not about the literal difference between 'binding' and 'mapping', but the difference between decoupling v4 & v6 and coupling v4 & v6. >cheers, >--satoru > _______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
