Re: Fix IPV6 literal notation?

[Brian E Carpenter <[e-mail]>]

On 29-Dec-20 07:57, Ofer Inbar wrote:
> On Sun, Dec 27, 2020 at 12:25:33PM +1300,
> Brian E Carpenter <[e-mail]> wrote:
>> On 27-Dec-20 11:16, Tim Bray wrote:
>>> See https://twitter.com/dave_universetf/status/1342685822286360576 to which I heartily concur. IPV6 addresses are neither easy for humans to read, nor easy for software to parse.
>>>
>>> *If* someone has a better idea, there???s no good reason not to standardize it, the old approach would still work.   Does anyone have a better idea?
>>
>> I have no idea if there's a proposal in that sequence of Twitter messages, but if there is, it should be written up as an I-D aimed at the 6man WG and discussed there. (Hence I have changed the IETF list to a Bcc: and added the 6man list in Cc:.)
>>
> 
> It's clearly not a proposal for changing anything, because the
> majority of the weirdness that thread describes is address forms that
> are no longer actively in use, but still appear in various places -
> old threads, documents, databases, config files, archives, etc.  Any
> change made to specifications now will only affect what forms people
> produce addresses in in the future, so it would have no effect on
> this.  IPv6 address formats are reasonably specified for the present
> and future, and changing specs won't affect artifacts of the past.
> 
>> I can say that this much is wrong:
>>
>>> Oh, and the leading zero debate also infects IPv6, to some extent! The specs tried to specify the textual representation of IPv6, but it failed to be complete. So it's unclear if 000001::00001.00002.00003.00004 is a valid IPv6 address
>>
>> It's quite clear that it is invalid. RFC4291 section 2.2 says:
>>
>>   "1. The preferred form is x:x:x:x:x:x:x:x, where the 'x's are one to
>>       four hexadecimal digits of the eight 16-bit pieces of the address."
>>
> 
> That's not clear at all, since it doesn't address validity, only what
> is "preferred". 

RFC4291 doesn't use RFC2119 language, but to me that sentence is a clear
specification. 02001:0:0:0:0:0:0:1 doesn't conform.

I also cited the ABNF, which is prescriptive for address literals in URIs,
but we all know that it applies generally.

> So if there is one thing in that thread that might
> call for a change, it's to clarify that IPv6 addresses are expected
> not to pad with leading zeros beyond 4 hex characters, and that it is
> invalid to write them with more digits.
> 
> I'd support that change, unless someone can show it's already very
> clearly spelled out somewhere.

I think that's really not needed. As a matter of curiosity, I think
the IPv4 format is pretty much undocumented. A few minutes of research
found this:

   One commonly used notation for internet host addresses divides the
   32-bit address into four 8-bit fields and specifies the value of each
   field as a decimal number with the fields separated by periods.  This
   is called the "dotted decimal" notation.  For example, the internet
   address of ISIF in dotted decimal is 010.002.000.052, or 10.2.0.52.

which first appeared in RFC820 (January 1983) and last appeared in
RFC990 (November 1986). Note that leading zeroes are allowed and do not
signal octal. 

    Brian