ietf
[Top] [All Lists]

Re: A simple question

2003-04-19 17:25:19
Thus spake "Robert Elz" <kre(_at_)munnari(_dot_)OZ(_dot_)AU>
| So if it's expected that both global and site-local addresses are
| available, why are we bothering with making things more complicated?

Because we need stable addresses for local use.

Any unique global address will work fine for local use, even if you're
disconnected from the Net.  Or do you expect your global prefix(es) to be
withdrawn from hosts every time your upstream connection flaps?

Something tells me you've never actually lived in an environment where
your global address changes moderately frequently.   If you had, you
wouldn't be so quick to ignore this need.

But haven't you heard?  With IPv6, renumbering will be so easy you'll do it
for fun!  </sarcasm>

The pain renumbering is one of the main motivators for NAT today, and I have
yet to see any reason to believe it'll be easier in IPv6.

  | Well.. all you need to do to fix this is to make a rule that if a
  | global prefix becomes available, the site-local prefix is no longer
  | appropriate and must be withdrawn.

Can't possibly work.

If site-locals and globals are to coexist, the obvious solution is to
specify that when the destination is SL, an SL should be preferred as
source, and likewise for globals.

Of course, that doesn't fix the case where you have multiple globals, or
multiple interfaces.  We haven't figured out how to fix that with IPv4
either, except for specific cases like "backside networks".

  | This *still* leaves the problem of using site-local behind a NAT,
though.

First, while I can imagine people existing who would be stupid enough to
do that, I find it hard to figure out what their reasoning would be.

Avoiding renumbering, avoiding exposing internal identities, etc.

It's all the same reasons you might use IPv4 NAT, with the possible
exception of address shortages.  I know several very large corps that use
legacy class B space inside and NAT to provider-assigned CIDR space
outside -- they're obviously not doing it because of address shortages,
they're doing it for administrative and scoping reasons.

I find it almost inconceivable to believe that anyone is deciding the fate
of SL addressing by reference to NAT - that's simply too ludicrous
(and sad) to contemplate.

Well, IPv6 NAT proponents expected to use the SL prefix as their private
addressing, so eliminating SL will require another prefix for that purpose.
I don't see any other direct connection, however.

S

Stephen Sprunk         "God does not play dice."  --Albert Einstein
CCIE #3723         "God is an inveterate gambler, and He throws the
K5SSS        dice at every possible opportunity." --Stephen Hawking




<Prev in Thread] Current Thread [Next in Thread>