Le 16/04/2019 à 22:48, Brian E Carpenter a écrit :
On 17-Apr-19 04:30, 神明達哉 wrote:
At Tue, 16 Apr 2019 12:58:01 +0200,
Alexandre Petrescu <alexandre.petrescu@xxxxxxxxx <mailto:alexandre.petrescu@xxxxxxxxx>> wrote:
Not quite, because it also says
" An Interface ID SHOULD be of length
64 decimal for all types of IPv6 addresses. In the particular case
of IPv6 link-local addresses, the length of the Interface ID MAY be
118 decimal."
which conflicts with RFC4291.
True. I forgot that 118. Thank you for pointing to it.
Remark, it says MAY, not MUST.
Do you stronly disagree with 118? I can remove the phrase containing
it, if so. I can also remove the entire cited text altogether, such
that to be silent about the length of the Interface ID.
(Speaking for myself who just happenned to notice it - I overlooked this
118, too). I'd say it's more consistent with the removal of "fe80::/10"
if we simply remove "In the particular case of IPv6 link-local
addresses..." sentence. If it really has to stay here, it will
inevitably need to be an update to RFC4291 and need to pass that high
bar (quite likely delaying the publication substantially, if not
making it fail). Unless that's absolutely necessary for this protocol
specification, it's much safer not to discuss that in this document.
Agreed.
I propose the following:
OLD:
A randomized Interface ID has the same characteristics of a
randomized MAC address, except the length in bits. A MAC
address SHOULD be of length 48 decimal. An Interface ID
SHOULD be of length 64 decimal for all types of IPv6
addresses. In the particular case of IPv6 link-local
addresses, the length of the Interface ID MAY be 118
decimal.
NEW:
A randomized Interface ID has the same characteristics of a
randomized MAC address, except the length in bits. A MAC
address SHOULD be of length 48 decimal. An Interface ID
SHOULD be of length specified in other documents.
Do you disagree?
Alex
Brian
