> There is a lot of similarity architecturally between net 10 in IPv4 and site
> local addresses
> in IPv6. Niether can be globally routed. Both spaces are re-used by large
> numbers
> of sites, and from casual inspection of a site-local address, outside the
> context of the inspection point, you can't tell what site it belongs to.
> And clearly people feel a need to use them or we wouldn't be
> having this discussion.
I think there are two motivations for site-locals (that not everybody
agrees with):
1. For sites that are isolated i.e. do not have a global prefix.
2. To completely isolate traffic local to a site from site-renumbering events.
I think most people agree with #1, but some think that #1 alone isn't
necessarily a good enough reason to have site-local addresses
in the architecture.
If we had a mechanism to handle #2 without using site-locals we could avoid
a lot of complexity. But handling #2, in my opinion, requires not only being
able to make long-running TCP connections survive renumbering, but also
somehow dealing with UDP (and TCP) applications which retain IP addresses
longer than the ttl indicated by DNS.
While fixing the applications might be easy in priciple, tracking down
all the places where an application might store an IP address for a longish
time, is quite difficult; much more difficult than the semi-automatic
porting of an application from IPv4 to IPv6.
Erik
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page: http://playground.sun.com/ipng
FTP archive: ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
