> > With this in mind, I see the following short term next steps: > > - Mark and I will produce a scenarios document that describes what > > situations we must solve > > - SOFTWIRE WG will be rechartered to add a work item on dual-stack > > lite/snat (token: SOFTWIRE chairs) > > - Progress the spec on dual-stack lite/snat (token: authors) > > - For the rest, we need a written analysis of the overall > > design space > > (token: Dan Wing and Alain Durand have agreed to do this) > > - Progress the specs for the different translation-based proposals, > > based on feedback in Dublin (token: authors) > > - More discussion is needed on implications of > > "carrier-grade NATs" in the pure IPv4 space > > > > Did I miss anything? > > Just a detail, but IMHO worth validating. > > Can it be assumed that the last but one item "Progress the specs for > the different _translation-based_ proposals..." is in practice open to > a proposal that avoids translation? > > The reason for the question is the proposal that is worked > on, following the APBP presentation in Dublin, with interest > expressed in particular by Teemu Savolainen and Gabor Bajko of > Nokia. > > With it, translation is completely avoided in scenario 1.d of Mark's > presentation (DS host accessing the IPv4-only world through > an IPv6-only cloud). > This transparency of global IPv4 across tunnels is naturally > seen as a positive point of the design.
I would consider that 'host-based translation', and hence in scope of what Jari mentioned. With host-based translation, the host is doing the translation instead of a device in the network. Depending on how it is implemented in the host, it could be the host application or the host's stack that 'borrows' the IPv4 transport address. I look forward to reading the I-D. -d _______________________________________________ Int-area mailing list [email protected] https://www.ietf.org/mailman/listinfo/int-area
