[cifs-protocol] 115102913315679 [MS-ADTS] 6.1.1.2.2.2.1 Subnet Object address range

[Douglas Bagnall]

hi Jeff,

> To follow up on this issue, how does this look?:
> 
> 2. The substring s[0, i-1] is either a valid IPv4 address in dotted decimal notation (as specified in [RFC1166]) or a valid IPv6 address in colon-hexadecimal form (as specified in [RFC2373]), and must meet the following additional constraints:  
> 	IPV4 addresses must not have any leading zeros in any individual component of the address.
> 	IPV6 addresses must not have any leading zeros in any individual component of the address.
> 	IPV6 addresses must be in compressed form (as specified in [RFC2373]) when possible; the compressed sequence must be the longest such possible.
> 	IPV6 addresses are treated as case insensitive

That is good, though the maximally compressed criteria doesn't quite
specify a unique form (which I believe is the point). For example, the
subnet that in uncompressed form looks like:

    1:0:0:2:3:0:0:4/128

can be compressed equally well in either of these two ways:

    1::2:3:0:0:4/128
    1:0:0:2:3::4/128

Windows 2012R2, as far as I can tell, mandates the first
representation (as does RFC5952 section 4.2.3).

Also in RFC5952 section 4.2.2 says that single zeros should not be
collapsed into "::" -- that is, it wants "2001:db8:0:1:1:1:1:1"
instead of "2001:db8::1:1:1:1:1". Windows 2012R2 also seems to follow
this, which is strictly speaking a violation of the compression
heuristic.

It looks as if Windows 2012 does in fact follow the RFC with the
caveat of case insensitivity. But not 2008R2, which allows invalid
subnet names like "a:b::/31", where the masked bits are not all zero
(online tools differ on whether that would refer to the same range as
"a:a::/31" or "a:b::/32", but either way its a duplicate).

thanks for your help,

Douglas