Re: Fix IPV6 literal notation?

[Alexandre Petrescu <[e-mail]>]

Le 27/12/2020 à 03:28, Karl Auer a écrit :
> On Sat, 2020-12-26 at 19:42 -0500, Gyan Mishra wrote:
>> I agree just on difficult to read and hard to parse.
> 
> Anything new is hard to start with and becomes easy with practice.
> 
> For a way to make it even easier, see RFC1924 ( 
> ftp://ftp.rfc-editor.org/in-notes/rfc1924.txt)
> 
>> The  :: Os compression helps, however does add confusion
> 
> I generally make the same recommendations as RFC5952.
> 
> The other thing people get hung up on is that the term "leading
> zeroes" does not mean ALL the zeroes. In most classes there is at
> least one person who tries to turn ":3000:" into ":3:" (and in one
> memorable case ":3001:" into ":31:").

YEs, that led to some confusion here too.

Another thing that leads to confusion is this fc00- and fd00- both being
a prefix for ULAs, being distinguished by just a prefix length.  One
cant say that all ULAs are fc00- prefixed addresses (like one says that
all LLs are fe80- prefixed addresses), because there are also the fd00-
prefixed addresses that are also ULAs.

Another problem with the IPv6 notation is the difficulty in memorizing
long strings such as 2001:2425:bcd:23:2DF:1DF:... When building an
addressing architecture one would rather choose hextets that are easy to
memorize.

It is true that there is a tendency to make work with addresses
dynamically changing all the time, but I have not yet seen an addressing
architecture plan that does not follow a certain logic understood by the
designer and that is not designed by hand.  I think it is rather
impossible to design an addressing architecture using random number
generators.  This is an additional reason why notation is useful.

Alex

> 
>> Understanding that when an IPv6 prefix is written that the leading 
>> higher order hex 0s are the not displayed within the hex field
>> which leads to confusion.
> 
> Can we blame IPv4 octal representation? There are no higher-order 
> zeroes shown in IPv4 octets either, and nobody gets confused.
> 
> A lot of people treat IP addresses as symbols without understanding 
> their underlying meaning. Making the link between the written
> addresses and the underlying bits helps a lot.
> 
> 128 bits are just plain harder to get your head around than 32. IPv6 
> addresses are not easily parsed by the unaided human; they are too
> big to be encompassed at a glance, regardless of the notation.
> 
> Regards, K.
> 
>