Re: [v6ops] A proposal for draft-ietf-6man-rfc4291bis-07 - LL notation

[Alexandre Petrescu <[e-mail]>]

Le 08/03/2017 à 04:28, Mark Andrews a écrit :
>
> In message <[e-mail]>, Alexandre Petrescu writes:
>> Thanks to this discussion, I would like to promote the following fact
>> that I did not know:
>>
>> If you say "FE80::" without telling the prefix length, that actually
>> means "FE80::/9" (and not FE80::/10, nor FE80::/64).  Because 0xFE80  is
>> "1111 1110 1000 0000" in binary representation.
>>
>> Alex
>
> No.
>
> FE80:: is always 128 bits.
> FE80::/9 says take the first 9 bits from those 128 bits.

YEs, and so it is if we just say "prefix FE80::" without specifying a plen.

> FE80::/10 says take the first 10 bits from those 128 bits.

YEs, and thus FE81::, FE82:: ... FEBF:: are all LL addresses.

I wonder why people only say FE80:: when talking LL address?

> FE80::/64 says take the first 64 bits from those 128 bits.
> FE80::/128 says take the first 128 bits from those 128 bits.
>
> All link local addresses have a prefix of FE80::/10.

But that does not mean that the notation of all LLs must start precisely 
with "FE80".

These notations could start with "FE81", "FE82" ... "FEBF".  Up to 
"FEC0" which is a GUA not an LL.

As such, a better way would be to say that LL addresses are of the form 
"FE##" where ## range from 81 to BF hexa.

> Link local addresses on ethernet are further restricted to
> have a prefix of FE80::/64.  They are a strict subset of
> all valid link local addresses.

YEs, and moreover, in this particular LL case, I suppose not only 
IP-over-Ethernet requires FE80::/64 but also SLAAC.  I.e. SLAAC on _any_ 
link (be it Ethernet or any other future link) seems to require the use 
of FE80::/64 prefix and does not allow for "FE81" or "FEBF".

An I-D describing the formation of LL addresses out of "FE##:://10" 
prefixes could make sense.

Alex