hi,
On 10/19/2011 6:21 AM, Maoke wrote:
hi Jacni,
在 2011年10月18日 上午11:54，Jacni Qin <[email protected]
<mailto:[email protected]>>写 道：
1) list all mutually exclusive IPv4 networks to be involved in
the residual deployment, i.e., Rule IPv4 prefixes;
2) calculate IPv6 prefix length for each Rule IPv4 prefix,
For clarifications, you mean the "Rule IPv6 Prefix" length, right?
since the length of Delegated Prefix should have been settled
before this step.
sure. i meant the Rule IPv6 Prefix length. :)
Ok, good.
there is still another issue, the MAX PSID. personally i like the
algorithm of making longest-match for the PSID. however, the position
of the PSID is somehow a problem. in the step 2) (or a little before),
when you setting the delegated prefix length, this length only covers
the IPv4 suffix and the PSID itself, right?
The Max PSID idea has been dropped by restoring the parameter "L", in
each rule.
Cheers,
Jacni
then if the length is, e.g., /64, and the PSID occupies the last 4
bits. then the MAX PSID for a CE is put at 60 ~ 71 bits. this is fine
with the routing, which can get the encapsulated stuff to the right
CE. however, it may hurt the IID to some extent, for example, we have
the PSID=0x4, while for a certain packet, it contains destination port
0xe488, then the MAX PSID = 0x488, and then the IID part for this
packet towards the dst CE will be not 200:5efe:: any more but
8800:5efe::
right? making the "u" bit destroyed, and the destination CE cannot
identify this is a packet for encapsulation or a packet for direct
forwarding. it also detroys the V-bit in the unified address format.
thus, may i understand that the planning should accommodate the MAX
PSID in the first 64 bits as well? if so, can we really satisfy the
requirement R-1 at the Beijing interim, stating "Efficient bit
representation. Address + Port range in =/56", if the PSID is less
than 4 bits?
sorry if i made anything, wrong. just try to understand the design,
its strength and limitations.
best,
maoke
Cheers,
Jacni
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires
