Re: [v6ops] draft-ietf-v6ops-ula-usage-recommendations - work or abandon?

[Fernando Gont <[e-mail]>]

El 9/11/2015 11:44, "Fred Baker (fred)" <[e-mail]> escribió:
>
>
>> On Nov 9, 2015, at 11:27 AM, Fernando Gont <[e-mail]> wrote:
>>
>> layer 3 addrs in layer 7 protocols is not clean design. Period.
>
>
> To my way of thinking, this is very fundamental. RFC 3439 makes a point
about coupling, and how coupling causes problems. When an application
presumes that an address that is meaningful in its environment is
meaningful in another, that is a very basic coupling. If the remote system
were given a name - an L7 identifier - that could be translated into an
address in the remote environment, which would resolve many of the issues.
The problem is not that NAT broke IPv4, or that IPv4 is broken; it is that
the applications written using IPv4 made an unfortunate coupling which NATs
exaggerate.
>
> I have a similar comment about the socket layer. Having the application
ask the OS for a translation and then connect to the address has been very
unfortunate for IPv6 deployment; had the socket layer been designed to have
the application present a character string (an address in textual form or a
name) and return a connected socket, IPv6 deployment would have been
invisible to the applications. It's a bad design at the socket layer, and
it's not IPv4's fault or IPv6's.

+1 (100% :-) )

Fernando