Hi, Nejc,
于 2011/8/31 7:23, Nejc Škoberne 写道:
Dear authors,
I have some comments and questions regarding draft-xli-behave-divi-pd:
1. If I understand correctly, each CPE is delegated a prefix of length
56 to 64 bits. The CPE can then do SLAAC for the network behind it, using
any of the subprefixes of the delegated prefix. Is this correct?
Almost.
The length of the delegated prefix for each CPE is from X to 64, where
X can be smaller than 56.
The CPE can do SLAAC for the network behind it, using any of the
addresses except the IPv4-translatable address and the CPE interface
address.
2. What IPv6 address is/should be configured on the CPE in the access
network? Can this be any IPv6 address?
This can be any IPv6 address except the IPv4-translatable address.
3. To me it seems that there is redundant information in the address
format - CPE index is encoded in the prefix (as "k" bits of the CPE index)
and in the suffix (as k). If I understand correctly, similar is true for
IPv4 address - the "s" bits in the CPE index uniquely identify one of
the shared IPv4 addresses. What exactly is the motivation for this? I mean,
IPv4 address and the port range are not encoded in the prefix anyway,
so direct CPE-CPE paths are not possible? I miss packet flow diagrams.
Not really. The embeding of the whole IPv4 address in IPv6 address
(i.e. IPv4-translatable address) is mandatory for IPv4/IPv6 stateless
translation as defined in RFC6052. The suffix which contains the
multiplexing
ratio and host index are required and also discussed in RFC6052. These two
components are the fundation for address-sharing stateless IPv4/IPv6
translation,
no matter it is single translation (IVI), dual translation (dIVI) or
dual translation
with prefix delegation (dIVI-PD). The CPE index only exists in prefix
delegation scenario, it contains the partial information of the IPv4
address
and the partial informationof the parameters of the address sharing. The
CPE index
is only used to identify the CPE, as the name states.
For dIVI-PD, direct CPE-CPE paths are supported.
4. I can't understand how exactly is "IPv4-converted address block"
supposed to be used. I really miss packet flow diagrams here. Also, you
seem to mix IPv4-converted and IPv6-converted.
The IPv4-converted address is for presenting the IPv4 hosts in the IPv4
Internet, which is defined in RFC6052.
"IPv6-converted" is a typo error, it should be IPv4-converted.
5. There is an example in Appendix A, which I cannot completely understand.
I cannot see how "v" (is this "s" in the CPE index?) is represented in
the IPv4-translatable address?
In the appendix we use v and h represent the IPv4 subnet (hex) and host
index (hex). The length of v is (s), and the length of h is (k) as shwon
in Figure 2.
Since you claim you have been running an implementation in production
successfully, I assume you have all these things sorted out. But I think
that the draft needs some more work to be understandable to others.
The dIVI has been running for two years and presented in IETF 76 (Nov.
2009).
The dIVI-PD have been running for one year.
The https://datatracker.ietf.org/doc/draft-xli-behave-divi/ and
https://datatracker.ietf.org/doc/draft-xli-behave-divi-pd/ should be read
together.
Regards,
xing
Thanks,
Nejc
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires
