Re: About IPv6 private address

[Remi Denis-Courmont <rdenis@simphalempin.com>]

On Tue, 5 Feb 2008 10:19:30 +1300, Nathan Ward <v6ops@daork.net> wrote:

> Longer term, it seems as though applications should support link local

> addresses. Has there been any documentation that disagrees with that?

> Safari reports link local addresses with the interface specified as

> being invalid (ie. %en1 at the end = 'invalid').



One complaint I have heard several time is, the notation is merely a de

facto standard extension of getaddrinfo et al. Also, as far as I know,

Windows does not implement it and for a reason:

fe80::216:41ff:fee1:b7fd%"Network Adapter" would not be very convenient and

fe80::216:41ff:fee1:b7fd%1 (if 1 is the scope ID is not very meaningful).



In any case, while applications should try to support these addresses, we

need to keep in mind that:

- the scope ID depends is hardware- and OS-dependant, if not boot sequence

dependant, hence cannot be used in a "for-dummies" documentation,

- application that rely on inet_pton(), inet_ntop() - and there may be

valid reasons - are kinda screwed,

- the scope being system-specific, cannot be exchanged accross the network:

think HTML hyperlinks, SDP c= lines, FTP control connection... and most

importantly DNS AAAA queries, so they are off limited practical usability

in any case.



I really think a home gateway ought to provide a ULA prefix, at least if it

cannot get any valid prefix from its upstream.



-- 

Rémi Denis-Courmont

http://www.remlab.net